# Aimbedded > AI that works inside your business, not beside it. We embed an AI operator in your team a few days a week, backed by a bench of AI agents and specialist partners. We find the work AI should be doing, build it into the tools you already use, and report the hours it gives back every week. ## The operator model Not a tool. Not a consultant. An operator. Most AI tools go unused and most strategy decks never ship. We embed an operator in your team, backed by our agent bench and partners, to build AI that works and count the hours it returns. - Embedded operator: works inside your team on set days each week, learns how the work really gets done, and owns the rollout from first win to handover. - AI agent bench: a bench of AI agents and specialist partners that research, build and test in parallel behind every operator. - Measured outcomes: every workflow starts with a baseline. Each week you see the hours it returned. We expand what works and switch off what does not. ## How an engagement runs 1. Score: a nine question assessment of your team, tools and time drains returns an instant AI opportunity score. 2. Plan: a 30 minute working session with an operator to pick the two or three wins likely to pay back fastest. 3. Embed: local businesses get a fixed setup package. Larger teams get an operator on set days each week, reporting hours returned. ## Principles - Your data stays yours: we build inside the tools you already own. No new data silos, no lock in. - People approve, AI drafts: anything customer facing goes to a person for approval until you decide otherwise. - Measured, then expanded: every workflow has a baseline and a weekly count. Scale only what proves itself. ## Company Page: https://aimbedded.co/company/ Based in San Francisco. Works on site and remote. First automation live in week one. - Embedded operator: Works inside your team, on site or remote, learns how the work really gets done, and owns the rollout from first win to handover. - Automation engineers: Build and test the workflows inside your tools, with the approval steps and logging each one needs. - AI agent bench: Agents that research, draft, build and test in parallel, so every engagement has more hands than one person. - Security and data partners: Review access, data handling and vendor settings before anything touches live customer or company data. - Specialist partners: Called in for a specific system or industry when the work needs deeper expertise than the core team. ## Tools we build in - Email and chat: Gmail, Outlook, Superhuman, Slack - Docs and data: Google Workspace, Microsoft 365, Notion, Airtable - Sales and CRM: HubSpot, Salesforce, LinkedIn - Projects and engineering: Asana, Jira, Linear, GitHub - Meetings: Fireflies, Granola, Fellow - Shops, wineries and restaurants: Shopify, Square, QuickBooks, Toast, Commerce7, WineDirect ## Common questions Q: How fast will we see results? A: Your first automation is live in week one. It starts in draft mode, with a person approving its output, and we report the hours it returns every week. Q: How is pricing set? A: Local businesses get a fixed setup package with a light monthly check in. Growing companies pay a monthly retainer based on how many days the operator is embedded. Enterprise work starts with a 90 day pilot in one department. Scope and price are agreed after the first working session. Q: Is there a minimum commitment? A: One month. Enterprise engagements start with a 90 day pilot. After the main work, support levels vary to fit what you need. Q: Who owns what you build? A: You do. Everything we build lives in your accounts, and there are no licence fees after the engagement. We keep our own reusable methods and templates. Q: What happens to our data? A: We work inside your accounts with the least access the work needs. We keep only anonymized patterns about how work gets done, never your business data, and we remove our access when the engagement ends. Q: What happens when an automation makes a mistake? A: Every automation starts with a person approving its output. Mistakes are logged and fixed the same day, the workflow pauses until the cause is fixed, and every error appears in your weekly report. Q: Is there work you do not take on? A: For now we do not handle patient health records or payment card data. Workflows are designed to route around them. ## Aimbedded for local businesses (wineries, shops, studios) Page: https://aimbedded.co/for/local-business/ A fixed setup package: a free audit, a one time AI setup, then a light monthly check in. Q: Is AI worth it for a business our size? A: Often, yes, when the same questions, bookings or orders take hours each week. We start with a free audit, so you see where the hours go before you commit to anything. Q: Do we need new software? A: Usually not. We build inside the tools you already use, such as Shopify, Square, QuickBooks, Toast, Commerce7, WineDirect, Gmail or Google Workspace. Q: How much of our time does it take? A: A short audit conversation, a little time to approve drafts in the first weeks, and a light monthly check in after setup. Problems we solve: - You answer the same customer questions by email, phone and DM, every day. - Bookings, club members or orders live in three tools that do not talk to each other. - Marketing happens when there is time, and there rarely is. - You know AI could help but have no time to work out where to start. Start: https://aimbedded.co/quiz/?segment=local-business ## Aimbedded for growing companies (20 to 500 people) Page: https://aimbedded.co/for/growing-company/ An embedded operator on set days each week who finds the highest value AI work and ships it with your team. Q: How is this different from hiring an AI engineer? A: You get an operator who works inside your team plus a bench of AI agents and specialist partners behind them, without the cost and wait of a full time hire. You can scale the days up or down month to month. Q: What do you work on first? A: The two or three workflows that return the most hours with the least risk, picked in a 30 minute working session. Common starting points are lead routing, weekly reporting, inbox triage and month end close. Q: Will our team actually use it? A: Adoption is part of the job. The operator works next to the people doing the work, trains them in their own tools, and only scales what the weekly numbers prove. Problems we solve: - You pay for AI seats, but use is patchy and nobody owns adoption. - Reports, handoffs and admin still get copied by hand between tools. - Software spend keeps rising while the work stays the same. - A full time AI hire is hard to justify yet. Start: https://aimbedded.co/quiz/?segment=growing-company ## Aimbedded for enterprise teams (500+ people) Page: https://aimbedded.co/for/enterprise/ An embedded operator inside one department at a time, turning pilots into working rollouts with measured savings. Q: How does an enterprise engagement start? A: With a 90 day pilot in one department. We agree a baseline, ship the first automation in week one, and report hours returned every week so the rollout decision rests on data. Q: Do you work with our IT and security teams? A: Yes. We work inside accounts you provision, with the least access the work needs, and we review data handling and vendor settings with your teams before anything touches live data. Q: Can you help with AI tools we already pay for? A: Yes. Mapping which licences are used, which overlap, and which work AI could take over is often the first piece of an enterprise pilot. Problems we solve: - Pilots stall between the innovation team and the people doing the work. - Platform licences overlap and nobody has mapped what AI could replace or retire. - Security and change management slow every rollout. - Leadership wants measured savings, not more demos. Start: https://aimbedded.co/quiz/?segment=enterprise ## Free AI score Page: https://aimbedded.co/quiz/ Nine questions, about three minutes. Returns an AI opportunity score from 0 to 100 and two or three concrete wins, then an optional free 30 minute working session with an operator. ## AI for AI seat adoption Page: https://aimbedded.co/use-cases/ai-for-ai-seat-adoption/ Industry: Enterprise IT. Task: Getting paid AI seats used and mapping overlapping software. Updated: 2026-10-10. Q: How can an enterprise get its paid AI seats used and find overlapping software? A: An enterprise can get paid AI seats used by embedding someone inside one department to find the tasks AI should handle, build them into the tools people already use, and measure real use each week. The same work maps which software licences overlap with what AI now does, so leaders can decide what to keep, consolidate or retire. Estimate: 100 to 250 staff hours a month. Based on typical setups, not a client result. ## The problem Many enterprises bought AI seats for whole departments. A few months later, usage reports show the same pattern: a handful of enthusiasts, a long tail of people who tried it once, and nobody who owns adoption. Leaders are asked to renew without evidence that the seats changed how work gets done. The gap is rarely the technology. People do not have time to work out which of their tasks AI should take on, and the examples in training sessions do not match their actual workflows. Pilots stall between the innovation team and the people doing the work. At the same time, software spend keeps rising. Departments run several tools that overlap: two note takers, three project trackers, separate form, survey and reporting products. Some of that work could now be handled by the AI seats already paid for, but nobody has mapped it. ## What the AI setup does 1. **Maps real work.** The operator interviews the department and reviews where time goes, using meeting notes, ticket and task data, and calendars, to list recurring tasks by hours spent. 2. **Picks the right tasks.** Tasks are ranked by hours, repetition and risk. The top few become the first builds. 3. **Builds inside existing tools.** Each task gets a working setup in Microsoft 365 or Google Workspace, Slack, Salesforce, HubSpot, Jira, Asana or Notion, so people use AI where they already work. 4. **Coaches by task.** Short sessions show each team how to use the new setups for their own work, with written guides kept in your shared space. 5. **Measures use weekly.** It tracks which setups are used, by how many people, and the hours they return, and flags teams where use drops. 6. **Maps overlapping software.** It builds an inventory of the department's tools, what each is used for, how often, and where it overlaps with another tool or with what AI now covers. 7. **Prepares the decision pack.** Leaders get a plain summary of options to keep, consolidate or retire, with the evidence behind each. ## Tools it works with We work inside your existing estate: Microsoft 365 or Google Workspace, Slack, Salesforce or HubSpot, Jira, Asana, Notion and Airtable, plus meeting tools such as Fireflies, Granola or Fellow where the department already uses them. We use the AI seats you already pay for, and access is granted and scoped by your IT team. ## How we would do it Enterprise work starts with a 90 day pilot in one department. An embedded operator joins that department, on site and remote, backed by a bench of AI agents and specialist partners. The first weeks set a baseline of current seat usage, time spent on recurring tasks and the tool inventory. Security and change management are planned in from the start with your IT team, rather than at the end. The first automation is live in week one, chosen so people see a real task get easier straight away. Over the pilot we add setups task by task and report adoption and hours every week. At the end of the 90 days, leadership gets measured savings, adoption figures and the overlap map, plus a plan for the next department. Our estimate for a department of a few hundred people is 100 to 250 staff hours a month returned, depending on how many recurring tasks move into working setups. ## What stays in your hands You approve everything. Every build is agreed with the department lead, anything customer facing is drafted by AI and approved by a person, and all software decisions remain with your leaders. You own what we build. Setups, guides and the software map live in your accounts, with no licence fees after the engagement. We keep only anonymized patterns about how work gets done, never your business data. Mistakes are logged and fixed the same day. The affected workflow pauses until the cause is fixed, and every error appears in the weekly report. Q: We already ran AI training. Why is this different? A: Training shows people what AI can do in general. We sit with one department, pick its most repetitive tasks, and build working setups inside its tools, so use comes from real daily work rather than good intentions. Q: Do you recommend which software to cancel? A: We map where tools overlap with each other and with what AI now handles, and show the usage evidence. Your IT, finance and department leaders decide what to keep, consolidate or retire. Q: Which AI product do you use? A: We work with the AI seats and tools you already pay for wherever possible. The goal is to get value from what you have before adding anything new. Q: How is adoption measured? A: We agree measures at the start, such as weekly active use by task, hours returned per workflow and tasks moved out of manual steps, then report them every week of the pilot. ## AI for inbound lead routing Page: https://aimbedded.co/use-cases/ai-for-inbound-lead-routing/ Industry: B2B services. Task: Inbound lead routing (HubSpot or Salesforce). Updated: 2026-10-10. Q: How can a B2B services company use AI to route inbound leads in HubSpot or Salesforce? A: A B2B services company can use AI to read each inbound lead, enrich it from the form, email and company website, score fit against your criteria, and assign it to the right owner in HubSpot or Salesforce within minutes. Reps get a short summary and a drafted first reply to approve, so good leads stop waiting in a shared queue. Estimate: 15 to 30 staff hours a month. Based on typical setups, not a client result. ## The problem Inbound leads are the cheapest pipeline you have, and they are often handled the slowest. Forms, emails to a shared inbox, chat transcripts and referral introductions arrive in different places. Someone has to read each one, look up the company, decide whether it fits, and pick an owner. When that person is busy or out, leads sit. Routing rules in HubSpot or Salesforce help, but they depend on clean fields. A form that says "Company: n/a" or a vague message like "can we talk about a project" defeats most rules. The result is a mix of leads sent to the wrong rep, good leads left in a general queue, and reps spending time on prospects who were never a fit. Leadership feels it too. Nobody is quite sure how long leads wait, why some convert and others do not, or whether the rules still reflect how the team sells. ## What the AI setup does 1. **Collects every inbound source.** Form fills, shared inbox emails, chat transcripts and referral notes all land as records in your CRM, with duplicates matched to existing contacts and companies. 2. **Reads and enriches the lead.** It reads the message and the company website, and checks public company pages such as LinkedIn, to fill in industry, size, location and the service the lead is asking about. 3. **Scores fit with a reason.** Each lead is scored against criteria you agree, and the score comes with a one line explanation a rep can check at a glance. 4. **Assigns an owner.** Using your territories, service lines and capacity, it sets the owner in HubSpot or Salesforce and notifies them in Slack with a short summary. 5. **Drafts the first reply.** The owner gets a drafted reply in Gmail or Outlook that references what the lead asked for and suggests times to talk. The rep edits and sends. 6. **Watches for stalls.** If a lead has no response after a set time, it nudges the owner, then the sales lead. 7. **Reports weekly.** A weekly summary shows volume by source, time to first reply, fit scores and overrides. ## Tools it works with This runs inside your existing CRM, HubSpot or Salesforce, with Gmail or Outlook for replies, Slack for alerts, and Google Workspace for the criteria sheet and weekly report. LinkedIn company pages help with enrichment. We do not add a new routing product or a separate database. Everything is stored on the records your team already uses. ## How we would do it Growing companies get an embedded operator on set days each week, paid as a monthly retainer by embedded days. The operator sits with your sales and marketing leads, maps how leads arrive today, and agrees the fit criteria and routing rules with the people who will live with them. In week one we switch on lead summaries in approval mode, so reps see what the AI would have said before it changes anything. Then we add scoring, assignment and drafted replies one step at a time, measuring time to first reply and rep overrides as we go. Each week you get a report of hours returned. Our estimate for a team handling a steady flow of inbound leads is 15 to 30 staff hours a month across marketing ops and sales, mostly from reading, research and data entry that no longer happens by hand. The minimum commitment is one month. ## What stays in your hands You approve everything. Fit criteria, routing rules and reply templates are agreed with your team, and no reply goes to a lead without a rep reviewing it. Reps can override any score or assignment in one click. You own what we build. Fields, workflows, prompts and reports live in your HubSpot or Salesforce and your other accounts, with no licence fees after the engagement. We keep only anonymized patterns about how this kind of work gets done, never your pipeline or customer data. Mistakes are logged and fixed the same day. If a lead is misrouted or a summary is wrong, the workflow involved pauses until the cause is fixed, and you see every error in the weekly report. Q: Does this replace our routing rules in HubSpot or Salesforce? A: No. It adds a reading and judgment step before your rules run. Simple territory or round robin rules stay in your CRM, and the AI fills in the fields those rules need. Q: What happens to leads that are a poor fit? A: They are tagged with a reason and sent to a nurture list or a polite decline draft, and a person can review that list each week to make sure nothing good slipped through. Q: Will reps trust the scores? A: Every score comes with a short plain reason, like company size, service asked about and budget signals. Reps can override any assignment, and those overrides are used to tune the criteria. Q: How fast is the first part live? A: The first automation is live in week one, usually lead summaries posted to Slack or the CRM record, before any routing changes are switched on. ## AI for month end close at a 200 person company Page: https://aimbedded.co/use-cases/ai-for-month-end-close/ Industry: Finance. Task: Month end close. Updated: 2026-10-10. Q: How can AI speed up month end close at a mid sized company? A: AI can match transactions, chase missing receipts, and draft variance explanations during month end close, so a finance team of three to six people closes days faster without adding headcount. Estimate: 20 to 40 finance hours a month. Based on typical setups, not a client result. ## The problem Close slips because people wait on receipts, match transactions by hand, and write the same variance notes every month. ## What the AI setup does 1. Matches bank and card lines to invoices and receipts, leaving only true exceptions for a person. 2. Messages employees with missing receipts automatically and tracks replies. 3. Drafts variance explanations from the ledger for the controller to edit. ## How we would do it An embedded operator on set days each week sits with the finance team, sets this up inside your existing accounting tools, and reports hours saved each month. ## Tools it works with QuickBooks or the accounting system you already run, your bank and card feeds, Gmail or Outlook for receipt requests, Slack for reminders, and Google Workspace or Microsoft 365 for the close checklist and variance notes. Nothing new for the team to learn. ## What stays in your hands Nothing posts to your ledger without a person approving it. The AI proposes matches, sends reminders you have approved the wording of, and drafts variance notes for the controller to edit. Your first automation is live in week one, usually receipt chasing, and it starts in draft and approval mode. You own everything we build, inside your own accounts, with no licence fees after the engagement. If an automation makes a mistake, it is logged and fixed the same day, the workflow pauses until the cause is fixed, and the error appears in your weekly report. Q: Does AI post entries to our ledger on its own? A: No. It proposes matches and drafts notes. A person on your finance team approves anything that changes the books. Q: Which accounting systems does this work with? A: We build inside the accounting and expense tools you already use, such as QuickBooks, plus your email and Slack for receipt chasing. Q: How soon does the close get faster? A: The first automation, usually receipt chasing or transaction matching, is live in week one. The full effect shows over the next one or two closes as the matching rules are tuned. ## AI for retail shop customer questions Page: https://aimbedded.co/use-cases/ai-for-retail-shop-customer-questions/ Industry: Retail. Task: Order and product questions (Shopify or Square shops). Updated: 2026-10-10. Q: How can a small retail shop use AI to answer order and product questions? A: A small retail shop can use AI to read each customer message, look up the order or product in Shopify or Square, and draft a clear reply about shipping status, sizing, stock, returns or pickup for staff to approve. The repetitive questions get answered fast and consistently, and your team only writes from scratch when a customer needs real help. Estimate: 8 to 15 staff hours a month. Based on typical setups, not a client result. ## The problem Most customer messages to a small shop are the same handful of questions. Where is my order? Do you have this in a medium? Can I return this if it does not fit? Is my pickup ready? Do you ship to my area? Each one means opening the inbox, finding the order in Shopify or Square, checking tracking or stock, and typing a reply. That takes a few minutes per message, and it happens dozens of times a week, usually while someone is also serving customers in the shop. Answers depend on who replies, so return windows and shipping times get described differently. Messages that arrive in the evening wait until the next day, and some shoppers buy elsewhere while they wait. Busy seasons make it worse. Order volume rises, so do questions, and the owner ends up answering emails late at night. ## What the AI setup does 1. **Sorts incoming messages.** It reads email and site contact form messages and labels each one: order status, product question, return or exchange, pickup, wholesale or press, or needs a person. 2. **Looks up the facts.** For order questions it finds the order in Shopify or Square and pulls status, items and tracking. For product questions it checks your catalog for sizes, materials and current stock. 3. **Drafts the reply in your voice.** Replies follow a short guide we write with you, covering your tone, return policy, shipping times and pickup rules, so every answer says the same thing. 4. **Flags anything sensitive.** Complaints, damaged goods, refund disputes and unusual requests go to a person with the order details attached and a suggested reply. 5. **Spots repeat questions.** A monthly note shows which questions come up most, so you can fix the cause, such as a vague product description or a missing size chart. ## Tools it works with The setup runs inside your existing accounts. It reads orders and products from Shopify or Square, works in your shop inbox in Gmail or Outlook, and keeps the reply guide and a simple log in Google Workspace. Nothing new for customers to learn, and nothing new for your team to log in to. We do not handle payment card data, and the setup never sees card details. ## How we would do it Local businesses get a fixed setup package. First, a free audit: we read a sample of recent customer messages, note how long each type takes, and show you which ones are worth handing to AI. Then a one time setup inside your accounts. We connect your shop and inbox, write the reply guide with you, and have the first automation live in week one, in approval mode. After setup comes a light monthly check in. We look at which drafts your team changed, update the guide for new products, sales or holiday shipping cutoffs, and report the hours returned. For a shop handling a steady flow of online orders, our estimate is 8 to 15 staff hours a month, depending on order volume and how many messages are simple status questions. The minimum commitment is one month. ## What stays in your hands You approve everything. Every reply starts as a draft that a person reads and sends. Refunds, discounts and exchanges are never issued by the AI. It prepares the information, and a staff member makes the call. You own the setup. The reply guide, automations and settings live in your accounts, with no licence fees after the engagement. We keep only anonymized patterns about how this kind of work gets done, never your customer or sales data. Mistakes are visible and fixed fast. If a draft is wrong, for example quoting an old return window, the error is logged and fixed the same day, and that workflow pauses until the cause is fixed. You see every error in the weekly report. Q: Can the AI see customer payment details? A: No. We do not take on work involving payment card data. The setup reads order status, items, shipping and product information only. Q: What if a customer is upset? A: Complaints, damaged items and anything that sounds unhappy are flagged for a person straight away, with the order details pulled up and a suggested reply your team can rewrite. Q: Does it work for both online orders and in store sales? A: Yes. Shopify shops and Square shops are both common, and many shops use both. The setup looks up whichever system holds the order. Q: Will it answer instantly at night? A: Replies start as drafts that a person approves. Once you are comfortable, you can let simple reply types, such as tracking links, send on their own outside opening hours. ## AI for support ticket triage Page: https://aimbedded.co/use-cases/ai-for-support-ticket-triage/ Industry: Customer support. Task: Ticket triage and draft replies in one department. Updated: 2026-10-10. Q: How can an enterprise support department use AI to triage tickets and draft replies? A: An enterprise support department can use AI to read each new ticket, tag its topic, urgency and product, route it to the right queue, and draft a reply from your help articles and past resolutions for an agent to approve. Agents spend less time sorting and searching, and more time on the cases that need judgment and care. Estimate: 150 to 300 agent hours a month. Based on typical setups, not a client result. ## The problem Large support teams lose a surprising share of their day before any customer is helped. Tickets arrive by email, web form and chat, often with vague subjects and missing details. Someone has to read each one, work out the product and issue, set priority, and move it to the right queue. Misrouted tickets bounce between teams and wait longer. Once a ticket reaches an agent, the next delay is searching. The answer usually exists somewhere: a help article, an internal runbook, a similar ticket solved last month, a known issue in Jira. Finding it and writing it up takes time, and quality varies between new and experienced agents. Many companies have already tried an AI pilot here. It often stalls between the innovation team and the support floor, because it was built beside the work instead of inside the tools agents use all day. ## What the AI setup does 1. **Reads every new ticket.** It reviews the subject, message, attachments list and customer record, and summarizes the issue in two lines at the top of the ticket. 2. **Tags and prioritizes.** It sets topic, product, language and urgency using your categories, and flags signs of churn risk, outage reports or legal language for fast human review. 3. **Routes to the right queue.** Using your routing rules and current queue load, it assigns the ticket and notes why. 4. **Links known issues.** If the ticket matches an open Jira issue or a recent incident, it links the two and tells the agent. 5. **Drafts a reply.** It writes a reply from your approved help articles, internal guides and past resolutions, citing the sources it used so the agent can check them. 6. **Learns from edits.** Agent changes to drafts are reviewed with your team leads each week, and the guidance is updated so the next drafts need fewer edits. 7. **Reports weekly.** Leaders see triage time, draft acceptance, routing accuracy and hours returned. ## Tools it works with The setup runs inside the systems your department already uses, such as Salesforce or HubSpot for cases and customer records, Jira for known issues and escalations, Slack for team alerts, and Gmail, Outlook or Microsoft 365 for email queues. Help articles and internal guides can come from Notion or your existing knowledge pages. No new agent console is added. ## How we would do it Enterprise work starts with a 90 day pilot in one department, here a single support team or region. An embedded operator joins that team, on site and remote, backed by a bench of AI agents and specialist partners. The first weeks set a baseline: how long triage takes, where tickets get misrouted, and which ticket types are most repetitive. Security and IT review happens up front, so access is agreed before anything connects. The first automation is live in week one, typically ticket summaries and tags in approval mode. Routing and drafted replies follow once the team trusts the tags. At the end of the 90 days, you get a measured result against the baseline and a plan for the next department, or for widening the same one. Our estimate for a department handling a high, steady ticket volume is 150 to 300 agent hours a month, depending on ticket mix and how much of the knowledge base is current. ## What stays in your hands You approve everything. Categories, routing rules and reply guidance are agreed with your support leaders, and every customer facing reply is drafted by AI and approved by an agent. You own what we build. Workflows, prompts, fields and reports live in your accounts, with no licence fees after the engagement. Support levels after the engagement vary by client. Mistakes are logged and fixed the same day. If a ticket is mistagged or a draft is wrong, that workflow pauses until the cause is fixed, and your leaders see every error in the weekly report. Q: Does the AI reply to customers on its own? A: Not during the pilot. Every reply is drafted by AI and approved by an agent. After the pilot, your team decides whether any narrow reply types can send without review. Q: How do you handle security review? A: We work inside your own accounts with access your IT and security teams grant, scoped to the one department in the pilot. We go through your review process before anything connects, and we keep only anonymized patterns about how work gets done, never your customer data. Q: Can it handle tickets with health or payment information? A: No. We do not take on work involving patient health records or payment card data. Those queues are excluded from the setup and stay with your existing process. Q: What do we measure at the end of the pilot? A: Time to first response, time to triage, how often agents accepted drafts with light edits, routing accuracy and agent hours returned, compared against a baseline taken in the first weeks. ## AI for weekly reporting Page: https://aimbedded.co/use-cases/ai-for-weekly-reporting/ Industry: Operations. Task: Weekly reporting from several tools. Updated: 2026-10-10. Q: How can a growing company use AI to produce weekly reports from several tools? A: A growing company can use AI to pull numbers and updates from the tools each team already uses, such as HubSpot, QuickBooks, Jira or Asana, check them against last week, and draft a weekly report with short explanations of what changed. A person reviews and sends it, so managers stop spending Monday mornings copying figures into slides and spreadsheets. Estimate: 20 to 40 staff hours a month. Based on typical setups, not a client result. ## The problem Most growing companies run on a weekly report that someone builds by hand. A manager logs in to the CRM for pipeline, the accounting system for cash and revenue, the project tool for delivery status, and a few spreadsheets for everything else. Numbers get copied into a template, charts get rebuilt, and someone writes a paragraph about why things moved. It takes a morning or more each week, often from people who should be running their teams. Small errors creep in: a filter left on, last week's number pasted twice, a column that changed meaning. Because the report is late or uncertain, leaders ask follow up questions in Slack, and the same people spend more time answering them. As the company adds tools and teams, the report gets longer and the work behind it grows with it. ## What the AI setup does 1. **Pulls the numbers on a schedule.** Each week it reads the agreed figures from your tools, for example pipeline and closed deals from HubSpot or Salesforce, revenue and cash from QuickBooks, and delivery status from Jira, Asana or Linear. 2. **Checks the data.** It compares each figure with last week and with the source system, and flags gaps, sudden jumps and numbers that disagree between tools. 3. **Drafts the commentary.** For each section it writes a short plain explanation of what changed and the likely reason, drawing on notes, tickets and deal updates. 4. **Builds the report.** It fills your existing template in Google Workspace, Microsoft 365 or Notion, keeping your layout and charts. 5. **Asks owners for the missing bits.** Where context is needed, it sends the right owner a short question in Slack and adds their answer. 6. **Routes for review.** The draft goes to the report owner, who edits and approves it before it is shared. ## Tools it works with We connect to what you already use. Common sources are HubSpot or Salesforce, QuickBooks, Jira, Asana or Linear, and Airtable or Google Sheets. The finished report lands in Google Workspace, Microsoft 365 or Notion, with a summary in Slack. The setup reads data with access you grant and keeps everything inside your accounts. ## How we would do it Growing companies get an embedded operator on set days each week, paid as a monthly retainer by embedded days. The operator starts by sitting with the people who build the report today. Together we list every figure, where it comes from, who owns it, and which ones leaders actually read. Usually some sections can be dropped. The first automation is live in week one, often a single section such as pipeline or delivery status, drafted alongside the manual version so you can compare them. We add sections one at a time until the full report is drafted automatically and reviewed by a person. Each week you get a report of hours returned. Our estimate for a company with several teams contributing to a weekly report is 20 to 40 staff hours a month, depending on how many tools and sections are involved. The minimum commitment is one month. ## What stays in your hands You approve everything. Every report is reviewed and approved by a named owner before it is shared, and you decide which figures and sources count as correct. You own the setup. The queries, templates, schedules and instructions live in your accounts, with no licence fees after the engagement. We keep only anonymized patterns about how reporting work gets done, never your financial or business data. Mistakes are treated seriously. A wrong figure or explanation is logged and fixed the same day, and that workflow pauses until the cause is fixed. You see every error in the weekly report. Q: What if our numbers disagree between tools? A: The setup flags the mismatch rather than picking one. A person decides which source is right, and that rule is written down so the same question does not come up again. Q: Can each team get its own version? A: Yes. The same data can feed a leadership summary and shorter team reports, each with only the sections that team needs. Q: Where does the report end up? A: Wherever your team reads it today, such as a Google Doc, a Notion page, a Microsoft 365 document or a Slack post. We do not add a new dashboard product. Q: Do the written explanations ever get things wrong? A: They can, which is why every report is reviewed by a person before it goes out. Corrections are logged, and the workflow pauses until the cause is fixed. ## AI for winery club renewals Page: https://aimbedded.co/use-cases/ai-for-winery-club-renewals/ Industry: Winery. Task: Club renewals. Updated: 2026-10-10. Q: How can a winery use AI to keep more wine club members renewing? A: A winery can use AI to spot club members likely to cancel, send each one a personal note before the next shipment, and answer their questions about the release automatically, so staff only handle the members who need a real conversation. Estimate: 3 to 6 staff hours a month saved, plus fewer cancellations. Based on typical setups, not a client result. ## The problem Most club cancellations arrive just before a shipment charges. By then the member has decided. Staff are busy pouring and rarely have time to call members who have gone quiet. ## What the AI setup does 1. Reads your club software export each week and flags members who skipped visits, changed shipping, or opened no emails. 2. Drafts a short personal note for each flagged member, using their past orders, for a staff member to approve in one click. 3. Answers common release questions (what is in the box, can I swap, when does it ship) by email and site chat. ## How we would do it A fixed setup package: a free audit of your club data and tools, a one time setup inside the software you already use, then a monthly check in to tune the notes and review results. ## Tools it works with Commerce7 or WineDirect for club data, Gmail for member notes, and Google Workspace for the weekly list your team works from. If you take tasting room payments through Square or Toast, purchase history from there can sharpen which members get a note. Payment card details are never touched. ## What stays in your hands Every note is approved by your staff before it goes out, in your voice. The first part, the weekly list of members at risk, is live in week one. You own the setup, which lives in your own accounts, with no licence fees after it is built. If something goes wrong, such as a note drafted for the wrong member, it is logged and fixed the same day, and that step pauses until the cause is fixed. The monthly check in reviews what was sent, who renewed, and what to tune next. Q: Will members know the note was drafted by AI? A: Every note is reviewed and sent by your staff, in your voice, using the member's own order history. You decide the tone and what gets sent. Q: Which club software does this work with? A: We build on your existing club and email tools, such as Commerce7 or WineDirect with Gmail, using the exports or connections they already offer. Q: Do we need to change how we run the club? A: No. The setup works around your shipment calendar and release process. Staff simply get a weekly list of members to contact and drafts to approve. ## AI for winery tasting room bookings Page: https://aimbedded.co/use-cases/ai-for-winery-tasting-room-bookings/ Industry: Winery. Task: Tasting room bookings and questions. Updated: 2026-10-10. Q: How can a winery use AI to handle tasting room bookings and visitor questions? A: A winery can use AI to answer the booking and visitor questions that arrive by email, phone message and site chat, offer open tasting slots from your calendar, and draft confirmations and changes for staff to approve. Your team stops retyping the same answers about hours, group sizes, pets and parking, and spends that time with guests in the room. Estimate: 6 to 12 staff hours a month. Based on typical setups, not a client result. ## The problem Tasting room questions arrive all day and from every direction. Someone wants to know if you take walk ins on Saturday. Someone else asks whether dogs are welcome, if there is a food option, or whether a party of eight needs a reservation. Another guest wants to move a booking from 2 pm to 4 pm. Each one is quick on its own, but together they pull staff away from guests who are standing at the bar. The answers usually live in three places: your booking tool, a shared calendar or spreadsheet for groups, and the heads of your longest serving staff. When the person who knows the policies is off, replies slow down or go out inconsistent. Messages that come in after hours wait until the next morning, and some of those guests book somewhere else. Peak weekends make it worse. The busier the room, the less time anyone has to check the inbox, so the people most ready to visit wait the longest for an answer. ## What the AI setup does 1. **Reads every new visitor message.** It watches your booking inbox, site chat and voicemail transcripts and sorts each message: simple question, new booking, change or cancellation, group or event request, or something that needs a person. 2. **Drafts answers from your own policies.** We write down your hours, tasting formats, group rules, pet and kid policies, accessibility details and directions in one short reference. The AI drafts replies from that reference only, so it does not make things up. 3. **Offers real open slots.** For booking requests it checks availability in your booking tool or calendar and suggests two or three open times in the draft reply. 4. **Prepares changes and cancellations.** It drafts the update in your booking system and the reply to the guest, and a staff member approves both with one click. 5. **Routes groups and events.** Larger parties and private event requests get a summary of the details collected so far, sent to your events lead with a suggested reply. 6. **Sends a short daily digest.** Each morning, staff see what came in overnight, what was drafted, and what is waiting on a person. ## Tools it works with We build this inside the tools you already run. Common setups use Commerce7 or WineDirect for reservations and club data, Gmail or Outlook for the booking inbox, and Google Workspace for a shared calendar, policy document and group request sheet. If you pour through Square or Toast at the tasting bar, we can read visit history from there to personalize replies for returning guests. We do not touch payment card data. ## How we would do it Local businesses get a fixed setup package. It starts with a free audit: we look at a sample of recent visitor messages, your booking flow and your policies, and show you which questions take the most staff time. Then comes a one time setup inside your existing accounts. We write the policy reference with you, connect the inbox and booking tool, and turn on the first automation in week one, in approval mode. After that, a light monthly check in. We review which drafts your team edited and why, tune the reference when policies or seasons change, and report the staff hours the setup gave back. Our estimate for a typical tasting room is 6 to 12 staff hours a month, depending on message volume and how many requests are groups. The minimum commitment is one month. ## What stays in your hands You approve everything. Every reply starts as a draft for a staff member to check, and you decide when, if ever, a simple reply type such as opening hours can send on its own. Bookings and changes are never made without a person confirming them. You own what we build. The automations, policy reference and settings live in your accounts, and there are no licence fees after the engagement. We keep only anonymized patterns about how this kind of work gets done, never your guest or business data. Mistakes are handled in the open. If a draft gets something wrong, it is logged and fixed the same day, and that workflow pauses until the cause is fixed. You see every error in the weekly report, along with what we changed. Q: Will guests know they are talking to AI? A: Yes, if you want them to. Most wineries add a short line to replies, and anything sent to a guest is drafted by AI and approved by a person until you decide a reply type is safe to send on its own. Q: Do we need new booking software? A: No. We build inside the booking, club and email tools you already use, such as Commerce7 or WineDirect with Gmail or Outlook. If your bookings live in a shared calendar or spreadsheet, we work with that. Q: What about large groups and private events? A: Those go to a person. The setup collects the details a host needs, like date, group size and occasion, then hands the request to your events lead with a draft reply ready to edit. Q: How soon does it start helping? A: The first automation is live in week one, usually drafted answers to your most common visitor questions, running in approval mode so your team checks every reply. ## AI consultant, agency, hire or tool Page: https://aimbedded.co/guides/ai-consultant-vs-agency-vs-hiring/ Q: Should I hire an AI consultant, an agency, a full time AI hire, or just buy a tool? A: It depends on how clear the problem is and who will own the work afterwards. Buy a tool when one task needs solving. Use a freelancer or agency for a defined build. Use a strategy consultancy for direction across a large organization. Hire full time when AI work is steady and core. Choose an embedded operator to find, build and measure the work inside your team. There are more ways to get AI working in a business than there were a few years ago, and each one suits a different situation. This guide compares six common options as fairly as we can, including the one we offer. The right answer depends on your problem, not on which option sounds most advanced. ## The six options at a glance | Option | Best for | What you get | Common risk | | --- | --- | --- | --- | | Buying an AI tool | One clear task that a product already does well | Software you set up and run yourself | It solves one task but does not connect to the rest of your work, and use fades | | Freelancer | A defined, small build with a clear spec | A specific automation or integration, delivered and handed over | Knowledge leaves with the person, and support after handover can be thin | | Automation agency | Several defined builds delivered on a project basis | A team that scopes, builds and ships a set of workflows | Builds are shaped by the project scope, so new needs mean a new project | | Strategy consultancy | Direction and planning across a large organization | Research, a roadmap, business cases and governance advice | Plans can stall between the strategy and the people doing the daily work | | Full time AI hire | Steady, long term AI work that is core to the business | A dedicated person who knows your business deeply over time | Hard to define, hire and keep the role, and one person cannot cover every skill | | Embedded AI operator | Finding and building AI work inside a team, then measuring it | A person working with your team on set days, backed by agents and partners | Needs time from your team, and works best when someone internal owns the result | ## Buying an AI tool This is the right choice more often than people think. If your biggest time sink is one clear task, and a well made product already handles it, buy the product. Many tools you already pay for, such as Google Workspace, Microsoft 365, HubSpot, Notion or Salesforce, now include AI features that are worth turning on first. It is a weaker choice when the real problem is the space between tools: information copied from email to a CRM, from a CRM to a spreadsheet, and from a spreadsheet into a report. A single tool rarely fixes that. ## Freelancer A good freelancer is fast and focused. If you can write down exactly what you need, for example "when a new order lands in Shopify, add the customer to our Airtable base and send a welcome draft for approval", a freelancer can often build it quickly. Choose this when the scope is small and well defined, and when someone on your team can look after the result afterwards. Plan for documentation and handover, because the knowledge otherwise stays with one person. ## Automation agency An agency brings a team, a process and more capacity than a single freelancer. It is a strong choice when you already know the workflows you want built and you want them delivered on a clear timeline. Agencies work best with a clear scope. If your needs are still unclear, spend time defining them before you sign, or the project may build the wrong things well. ## Strategy consultancy Large organizations often need help deciding where AI fits, how to govern it and how to explain the plan to leadership. A strategy consultancy is good at research, structure and building the case for change across many departments. The gap is usually between the plan and the work. Make sure someone owns turning the roadmap into working tools inside the teams that will use them. ## Full time AI hire When AI work is steady, central to how the business runs, and likely to grow for years, hiring is often the best long term choice. An internal person builds deep knowledge of your systems and people that no outside provider can match. The difficulty is timing. Before there is a clear body of work, the role can be hard to define, and one person rarely has every skill needed, from process mapping to building to change management. Many companies bring in outside help first and hire once the work is proven. ## Embedded AI operator This is the model we offer. An operator works inside your team, on site or remote, on set days each week. They find the work AI should be doing, build it inside the tools you already use, and report the hours returned every week. A bench of AI agents and specialist partners supports them, so you are not limited to one person's skills. It suits businesses that know AI could help but have not yet worked out where, and that want working results rather than a plan. It is not the right choice if you only need one tool switched on, or if you already have a strong internal AI team with spare capacity. ## How to decide Three questions narrow it down quickly: 1. **Is the problem clear?** If you can describe the exact task, a tool, freelancer or agency may be enough. If not, you need someone to find the work first. 2. **Who will own it afterwards?** Every option needs an owner on your side. If nobody has time, choose an option that includes ongoing support or plan to hire. 3. **How will you measure it?** Agree a baseline in hours before anything is built, whichever option you choose. Any savings figures should be treated as estimates and checked weekly. ## Where to go next The [free AI score](/quiz/) takes about three minutes and helps show whether your next step is a tool, a build or something more embedded. See how each approach plays out for [local businesses](/for/local-business/), [growing companies](/for/growing-company/) and [enterprise teams](/for/enterprise/), or look at concrete examples on the [use cases page](/use-cases/). Q: Can we combine more than one of these options? A: Yes, and many businesses do. A common mix is buying tools for simple tasks, bringing in outside help to set up the harder workflows, and later handing ownership to an internal person. Q: Is buying an AI tool enough for a small business? A: Sometimes. If one task takes most of the time and a tool already does it well, buying that tool is the simplest answer. It gets harder when the work spans several tools that do not talk to each other. Q: When does a full time AI hire make sense? A: When there is enough steady AI work to fill a role for years, and someone senior can manage and support that person. Before that point, the role can be hard to define and hard to keep. Q: What should we ask any outside AI provider before signing? A: Ask who owns what they build, what happens when they leave, how mistakes are caught and reported, and how they will measure the hours returned. Clear answers to those four questions matter more than any demo. ## How to choose your first AI project Page: https://aimbedded.co/guides/how-to-choose-your-first-ai-project/ Q: How do I choose the first AI project for my business? A: Start with the repetitive work your team already does every week. List it, then score each task by hours spent, risk if it goes wrong, how easy the data is to reach, and who needs to approve the output. Pick two or three low risk, high hour tasks, record how long they take today, and measure the hours returned every week. Most businesses pick their first AI project the wrong way. They start with the most exciting idea, or the one a vendor demoed, and then look for a place to use it. A better approach starts with the work your team already does, and picks the parts that are boring, frequent and easy to check. This guide walks through a simple method that works for a five person shop and for a department inside a large company. ## Step 1: List the repetitive work Spend one week writing down the tasks that repeat. Ask each person on the team to keep a short list. You are looking for work that follows a pattern, even a loose one. Good candidates usually sound like this: - Answering the same customer questions by email, phone or message. - Copying information from one tool into another, such as from Gmail into HubSpot or from Shopify into a spreadsheet. - Writing the weekly report, the meeting summary or the handoff note. - Chasing people for missing information, approvals or receipts. - Sorting, tagging or routing incoming requests. Do not filter yet. A long, messy list is fine. The scoring step does the sorting. ## Step 2: Score each task on four things Give each task a simple score from 1 to 5 on four questions. Keep it rough. You are comparing tasks with each other, not building a business case. **Hours.** How many staff hours does this take in a typical month, across everyone who does it? A task that takes ten minutes but happens forty times a week matters more than one long task each quarter. **Risk.** What happens if the output is wrong? A draft internal summary with a small mistake is easy to fix. A wrong price sent to a customer is not. Lower risk scores higher here. **Data access.** Does the information the task needs already live in tools you use, like Google Workspace, Microsoft 365, Slack, Salesforce or QuickBooks? If someone has to dig through paper files or ask three colleagues, score it lower. **Approval needs.** How many people need to sign off before the output is used? One person checking a draft is ideal. A task that needs legal, finance and a manager to approve each time will move slowly. Add the four scores. The tasks at the top of your list are your candidates. ## Step 3: Pick two or three wins Choose two or three tasks from the top of the list. Not one, because a single project can stall for reasons that have nothing to do with AI. Not ten, because nobody can watch ten new workflows at once. A good starting set usually has: - One task that saves a clear number of hours for one person or role. - One task that removes a handoff between two tools or two teams. - Nothing that sends anything to a customer without a person approving it first. Write down, in one sentence each, what done looks like. For example: "The weekly sales summary is drafted in Notion every Monday morning and the sales lead edits and sends it." ## Step 4: Set a baseline before you change anything This is the step most teams skip, and it is the one that tells you if the project worked. For each chosen task, record how it runs today: - How many times it happens in a week. - Roughly how long each one takes. - Who does it. - How often it goes wrong or needs rework. Estimates are fine. Ask the people who do the work. Their answer will be more accurate than a guess from a manager. ## Step 5: Start in draft mode and measure weekly Run each new workflow in draft and approval mode first. The AI prepares the work and a person reviews it before it goes anywhere. This keeps risk low while the team learns what the AI gets right and where it needs help. Each week, compare against your baseline: - How many times did the workflow run? - How long did review take compared with doing it by hand? - How many drafts needed real changes? - What went wrong, and was it fixed? Turn this into an estimate of staff hours returned. Keep the estimate conservative. If a task used to take 30 minutes and now takes 10 minutes to review, count 20 minutes, not 30. After four to six weeks you will know which projects are worth keeping, which need changes, and which to drop. Then go back to your list and pick the next two or three. ## How this changes by business size The method is the same at every size. What changes is the scope. **Local businesses** such as wineries, shops and studios usually find their first wins in customer questions, bookings and orders spread across tools like Square, Toast, Shopify, Commerce7 or WineDirect. One owner or manager can run the whole scoring exercise in an afternoon. A single well chosen task can return an estimated few hours a week. **Growing companies** of 20 to 500 people often have plenty of AI seats but patchy use. Here the best first projects tend to be reports, handoffs and admin that get copied by hand between tools like HubSpot, Asana, Jira or Slack. Run the listing step team by team, then score across the whole company. **Enterprise teams** should start inside one department, not across the whole business. Pick a department with a clear owner and a willing team. Bring security and change management in at the scoring stage, because approval needs often decide which tasks are realistic in the first 90 days. ## Where to go next If you want a quick read on where your business stands, take the [free AI score](/quiz/). It takes about three minutes and points to the areas worth scoring first. You can also see how this looks for [local businesses](/for/local-business/), [growing companies](/for/growing-company/) and [enterprise teams](/for/enterprise/), or browse worked examples on the [use cases page](/use-cases/). Q: How long should a first AI project take to show results? A: A good first project should show something working within a week or two. If it needs months of setup before anyone sees a result, it is probably too big to start with. Q: Do we need clean data before we start? A: No. Pick tasks where the information already lives in one or two tools you use every day, such as your inbox, calendar or CRM. Large data cleanup projects make poor first projects. Q: Should the first project be customer facing? A: Usually not. Internal work like reports, handoffs and admin is a safer place to learn. If you do start with customer replies, have AI draft them and a person approve every one. Q: What if none of our tasks score well? A: That is rare, but it happens. It usually means the work is spread across too many tools or depends on judgement calls. Start by fixing one handoff between tools, then score again. ## What an embedded AI operator does Page: https://aimbedded.co/guides/what-an-embedded-ai-operator-does/ Q: What does an embedded AI operator do? A: An embedded AI operator works inside your team, on site or remote, on set days each week. They find the work AI should be doing, build it inside the tools you already use, and report the hours returned every week. A bench of AI agents and specialist partners supports them, every automation starts in approval mode, and you own everything built. Most AI help arrives from the outside. A tool gets bought, a consultant writes a plan, or an agency ships a build and moves on. An embedded AI operator works the other way around. They join your team, learn how the work actually gets done, and build AI into that work alongside the people who do it. This guide explains what that looks like in practice. ## The basic idea An operator is a person who sits inside your team on set days each week, on site or remote. Their job has three parts: 1. **Find** the work AI should be doing, by watching how tasks move through your team and your tools. 2. **Build** it inside the tools you already use, such as Gmail, Outlook, Slack, Google Workspace, Microsoft 365, HubSpot, Salesforce, Asana, Jira or QuickBooks. 3. **Measure** the result, and report the estimated staff hours returned every week. The operator is not there to sell you new software. Most of the work happens in accounts you already pay for. ## What a week looks like No two weeks are the same, but most follow a similar rhythm. **Start of the week.** The operator reviews last week's report with the person who owns the engagement on your side. What ran, what was approved, what went wrong and what was fixed. Together they agree the priorities for the week. **Time with the team.** The operator sits with the people doing the work. They ask how a task is done today, where it gets stuck, and what would help. This is where the best ideas come from, and it is the part that tools and remote builds tend to miss. **Building and testing.** New workflows are built in your accounts and tested on real but low risk cases. Each one starts in draft mode, where the AI prepares the work and a person approves it. **Training and handover.** The operator shows the team how each workflow behaves, what to check, and how to pause it. The goal is that your people are comfortable running it without help. **Weekly report.** At the end of the week you get a short report: workflows running, estimated hours returned, errors found and how they were fixed, and what comes next. The first automation goes live in week one, so the team sees something working early rather than waiting for a long discovery phase. ## The bench behind the operator One person cannot be expert in everything. Behind each operator is a bench of AI agents and specialist partners. The **agents** handle the repetitive parts of the operator's own job: mapping processes, drafting documentation, testing workflows and preparing reports. This frees the operator to spend more time with your team. The **specialist partners** step in when a project needs deeper skills, for example a complex CRM setup or a tricky data connection between tools. You deal with one operator, and they bring in the right help when it is needed. ## Approval and mistakes Trust is earned one workflow at a time. Every automation starts in draft and approval mode. Anything customer facing is drafted by AI and approved by a person on your team before it goes out. You approve everything we build. Mistakes still happen. When one does, it is logged, fixed the same day, and that workflow pauses until the cause is fixed. You see every error in the weekly report, not just the good news. ## How it is measured Before a workflow is built, the operator records a baseline: how often the task happens, how long it takes and who does it. After it goes live, they track how long review takes and how often drafts need real changes. The difference becomes an estimate of staff hours returned. We keep estimates conservative and show our working, so you can check them against what your team sees. We report hours, not money, because hours are what your team can actually feel. ## How engagements are priced There are three pricing models, matched to the size of the business. Pricing depends on scope, so we share figures after a first conversation. - **Fixed setup package** for local businesses. A free audit, a one time AI setup, then a light monthly check in. - **Monthly retainer by embedded days** for growing companies. The retainer reflects how many days the operator spends with your team. - **90 day department pilot** for enterprise teams. One department at a time, with clear goals agreed at the start and measured at the end. ## What you provide This varies by client and by project. Most engagements need: - A person on your side who owns the engagement and meets the operator each week. - Access to the accounts and tools the work touches, set up the way your team prefers. - Time from the people who do the work, mostly at the start, so the operator can learn the real process. - Someone to approve drafts while workflows are in approval mode. Larger organizations will usually also involve security and IT early. A small business may need little more than an owner with an hour or two a week. ## What happens at the end You own everything we build in your accounts. There are no licence fees after the engagement ends, and nothing stops working because we leave. Before the end, the operator documents each workflow and makes sure your team knows how to run, pause and adjust it. Support after the engagement varies by client. Some businesses want a light check in, some want the operator to stay on for the next set of projects, and some prefer to run everything themselves. ## Where to go next The [free AI score](/quiz/) takes about three minutes and shows where an operator would likely start in your business. See how the model works for [local businesses](/for/local-business/), [growing companies](/for/growing-company/) and [enterprise teams](/for/enterprise/), or browse examples on the [use cases page](/use-cases/). Q: How fast does the first automation go live? A: The first automation is live in week one. It starts in draft and approval mode, so a person on your team checks its work before anything is used. Q: Will the operator see our customer and business data? A: We work inside your own accounts with the access you grant. We keep only anonymized patterns about how work gets done, never your business data. Q: Is there any work you will not take on? A: Yes. We do not take on work involving patient health records or payment card data. Q: What is the minimum commitment? A: One month for local businesses and growing companies. Enterprise engagements start with a 90 day pilot in one department. Savings figures on this site are estimates based on typical setups, not client results.