Plain-English guides to AI agents, AI CRM, automation and governance for small businesses: what each one is and how to start safely.
StartAI agents for small business
An AI agent is software that takes a goal, decides on steps and uses tools such as email, a calendar or a spreadsheet to carry them out. A chatbot answers a question; an agent finishes a task.
For a small business the useful first agents are narrow: answer common customer questions, draft follow-up emails, keep a lead list tidy, or prepare a weekly report. Start with one task you already do by hand every week, write down what a good result looks like, and keep a person approving the output until it is reliably right.
Related: How we work · Find customers
Back to all guides ↑StartAI CRM for small business
A CRM keeps one record per customer: who they are, what they bought, and every conversation. An AI CRM adds help on top of that record, such as summarising a thread, suggesting the next step or drafting a reply.
Pick a CRM for the record first and the AI second. Check that you can export your data, that it connects to the email and calendar you already use, and that every AI suggestion can be reviewed before it reaches a customer.
Related: Find customers · Run it
Back to all guides ↑StartAI CRM migration
Moving to a new CRM is mostly a data job. List every field you use today, decide where each one goes in the new system, and clean duplicates before you move anything.
Run the move on a copy first, then compare record counts and spot-check real customers on both sides. Keep the old system read-only for a while so nothing is lost, and switch any automations over only after the data checks out.
Related: Data · How we work
Back to all guides ↑BuildAI workflow automation
Workflow automation connects steps that people repeat: a form is filled in, a record is created, someone is told, a follow-up is scheduled. AI adds the steps that need reading or writing, such as sorting a request or drafting a reply.
Map the workflow on paper before you automate it. Automate the dull, rule-based steps first, put a human check on any step that sends something to a customer, and log every run so you can see what happened when something goes wrong.
Related: AI and machine learning · Cloud and DevOps
Back to all guides ↑BuildAI agent orchestration
Orchestration is how several agents work together: one plans, others research, write or check, and a coordinator passes work between them and decides when a job is done.
Good orchestration gives each agent one clear job, a defined hand-off, and a check before work moves on. Keep a record of who did what, so a person can trace any result back to the step that produced it.
Related: How we work · AI and machine learning
Back to all guides ↑BuildCRM integrations
An integration keeps your CRM in step with the other tools you use: website forms, email, calendar, invoicing and support. Without it, people retype data and records drift apart.
Decide which system owns each piece of data, sync in one direction where you can, and watch for failed syncs. Use the official connection a tool offers before building a custom one.
Related: Data · Web
Back to all guides ↑RunAI agent governance
Governance is the set of rules an AI agent works under: what it may do alone, what needs a person's approval, what it must never do, and how its actions are recorded.
Write the rules down, give each agent only the access it needs, require approval for anything that spends money or contacts a customer, and keep an audit trail. Review the trail regularly and tighten the rules when something surprises you.
Related: Run it · Terms
Back to all guides ↑RunAI autonomy levels
Autonomy levels describe how much an AI system is allowed to do without a person: from suggesting only, to acting with approval, to acting alone inside set limits.
What a system can do and what it is permitted to do are separate questions. Research on agent governance proposes treating capability and permission as separate settings (arXiv 2607.23438). Raise a system's permitted level one step at a time, and only after its record at the current level is good.
Related: Separating capability from permission (arXiv) · How we work
Back to all guides ↑RunCRM automation
CRM automation runs routine updates for you: moving a deal to the next stage, assigning a new lead, setting a reminder, or sending a scheduled follow-up.
Start with reminders and assignments, which are low risk, before automating anything customers see. Name each rule clearly, keep a list of what is switched on, and check it after any change to your sales process.
Related: Find customers · Run it
Back to all guides ↑GrowAI lead follow-up
Lead follow-up is answering a new enquiry quickly and then staying in touch until the person decides. AI can draft replies and suggest timing based on what the lead asked.
Keep follow-up honest and easy to stop: answer what was asked, respect unsubscribe requests at once, and have a person approve messages before they go out until the drafts are reliably good.
Related: Find customers · SEO and content
Back to all guides ↑GrowAI lead qualification
Qualification sorts leads by how well they fit what you sell, so time goes to the ones most likely to buy. AI can read an enquiry and a public company profile and suggest a fit score with its reasons.
Write your fit criteria down first, ask the AI to show the evidence for each score, and spot-check its choices. A score without reasons is hard to trust and hard to fix.
Related: Find customers · Get started
Back to all guides ↑GrowAI CRM pricing and ROI
To judge whether an AI CRM pays off, compare the time and money it costs with the time it saves and the sales it helps win. Count setup, data clean-up and training, not only the subscription.
Measure before you switch: how long replies take, how many leads go unanswered, how many deals close. Measure the same things after a few months. Vendor prices change often, so check each vendor's current pricing page rather than any figure quoted elsewhere.
Related: Sell online · Contact
Back to all guides ↑