AI for small business
An AI business assistant that works inside your existing tools
An AI business assistant is software that takes work off an owner's plate by understanding plain instructions — spoken or typed — and carrying them out across the tools a business already uses: calendar, email, job management, invoicing. Unlike a chat app that only answers questions, a business assistant executes: it drafts the follow-up, updates the record, schedules the visit, and reports back. The useful ones keep a person in charge of anything sensitive.
The jobs it can genuinely handle
Not everything — these are the repeatable, rule-bound workflows where this actually works today.
Turning voice notes into finished actions
A messy 40-second voice memo from the truck — 'job's done, collect the balance, they want a quote on the fence too' — becomes a drafted invoice, an updated job record, and a scheduled follow-up, each waiting for your OK.
Drafting messages in your voice, with context
Follow-ups, estimates cover notes, and replies are drafted using the customer's actual history — what was quoted, what was promised, what happened last time — instead of a blank compose window.
Keeping records current without data entry
Outcomes get logged, job statuses updated, and contact details corrected as work happens, so your systems stop depending on you catching up on admin at night.
Preparing your day before it starts
A short morning brief: today's schedule, the estimates going stale, the balances aging, the customer who went quiet — with a suggested next action on each that you can approve from your phone.
Answering questions about your own business
'What did we quote the Hendersons?' 'Who's overdue for their fall service?' Answered from your systems' records in seconds, instead of a search through three apps and a text to whoever might remember.
Watching for what needs attention
Stale quotes, calendar gaps, dormant customers, a review that needs a response — surfaced to you before they quietly cost money, rather than discovered when it is too late to act.
Chasing routine loose ends
Unsigned documents, unanswered messages, unconfirmed appointments — followed up politely until the loop closes, with a record of every attempt.
What it cannot safely decide
Anyone selling this without naming its limits is selling it badly. These calls stay with a person.
- What your priorities are. It can surface everything that needs attention; deciding what matters most this week is the owner's job.
- Commitments beyond your rules — new pricing, policy exceptions, promises about scope or timing. It drafts within your rules and asks when a request falls outside them.
- Judgment about people: hiring, performance, scheduling conflicts between staff. It can organize the information, never make the call.
- Legal, financial, or clinical decisions. It prepares and organizes; the decision belongs to you or your qualified professional.
- What you meant when an instruction is ambiguous. A good assistant asks a clarifying question; a dangerous one guesses and acts.
How Deptly draws this line in practice — approvals, escalation, and audit history — is covered on trust and control.
What it needs to work with
No system does this from a blank slate. Expect to connect or share some of the following.
- Access to the tools where the work actually lives — email, calendar, job or practice management, invoicing — with permission scopes you set per system.
- Your business basics: services, pricing rules, customers, and enough history that it recognizes a longtime customer as one.
- Your standing rules and preferences — tone, approval thresholds, who on the team handles what, which customers get special handling.
- A channel to reach you the way you actually work: voice notes, texts, or an app, for both instructions and approvals.
- Ongoing corrections. When you fix its work, the fix should become a rule it keeps — that feedback loop is part of the data.
Oversight options, from tightest to lightest
You choose how much runs on approval and how much runs on approved rules — and you can move up or down the ladder any time.
- Draft-only: everything is prepared — messages, record changes, schedule moves — and nothing happens until you say so.
- Approval queue: actions are batched for quick review, so ten decisions take one minute on your phone instead of ten interruptions.
- Approved routines: actions you have explicitly cleared — confirmations, record updates, standard follow-ups — run on their own; anything novel still asks first.
- Rules with escalation: the assistant works within written rules, and anything ambiguous, sensitive, or new comes back to you as a question, never as a completed surprise.
- A full activity log at every level — what it did, when, on whose instruction, and in which system.
What this looks like in a real business
Three examples in the vocabulary the industry actually uses.
Plumbing and HVAC
After a water-heater install, the tech's voice note becomes the drafted invoice, a permit note on the job record, and a follow-up scheduled for the aging water softener he spotted — all before the truck leaves the street.
Deptly for plumbing and hvacMed spa
The morning brief lists today's clients, the memberships due for renewal, and the two consults who never booked treatment. The owner replies 'follow up on both, usual offer' from her phone between appointments — clinical questions stay with the provider.
Deptly for med spaMoving and estate cleanout
A crew lead texts photos of the finished estate cleanout. The assistant drafts the final invoice against the estimate, logs the disposal receipts to the job, and queues the review request for the executor — the owner approves all three with one tap.
Deptly for moving and estate cleanoutWhat actually drives the price
Quotes vary widely because scope varies widely. These are the factors that move the number.
- How many systems it needs to work in, and whether each has a real integration or requires browser-based access.
- Volume: how many instructions, messages, and actions flow through it in a month.
- Depth of setup — how much business knowledge, history, and how many rules it starts with determines both onboarding effort and early usefulness.
- The oversight level you want; approval-heavy configurations trade your minutes for certainty.
- Software you operate versus a managed service that sets it up, tunes the rules, and keeps it current as your business changes.
For how Deptly scopes and prices this kind of work, see pricing.
How to evaluate any provider
Take this checklist into every sales conversation — including one with us.
- Can it take real action in your tools — send, schedule, update — or does it only produce text for you to paste somewhere?
- Give it a deliberately ambiguous instruction in the demo. Does it ask a clarifying question, or guess and act?
- What does it remember between sessions — customers, rules, corrections — and where is that memory stored?
- Can you read a complete log of every action, with the instruction and reasoning behind each?
- What is the escalation path when it hits something outside its rules — and how fast does that reach a person?
- When you correct its work, does the correction stick as a rule, or will you fix the same mistake monthly?
- Is your data stored securely, kept out of other customers' models, and exportable if you leave?
- Ask for a walkthrough on one of your real workflows — your actual tools, your actual customer scenario — before committing.
Where this lives at Deptly
Inside Deptly
Deptly's assistant layer is built from the pieces this page describes: business memory, owner decision memory, voice-to-execute, and approvals. Inside Deptly shows plainly how each works, what information it uses, and where human review applies.
Questions owners ask
A virtual assistant is a person working remotely; an AI business assistant is software. The person brings judgment and can handle genuinely novel situations; the software brings instant availability, perfect consistency on routine work, and direct access to your systems' records. Many businesses use both — the software handles the repetitive loops so the human's hours go where judgment matters.
A general chat tool knows nothing about your business and cannot touch your systems — every session starts blank, and everything it drafts, you still execute. A business assistant is connected: it knows your customers, prices, and history, acts inside your calendar and invoicing, and remembers your corrections. The difference is between advice and execution.
Pricing spans self-serve software subscriptions to managed services priced on scope. The drivers are the number of systems connected, monthly action volume, setup depth, and whether a team maintains it for you. Compare against the concrete alternative for your situation — part-time admin help, or the owner-hours currently going to the work around the work.
Basic drafting help works almost immediately. Real usefulness — knowing your customers, your prices, your rules well enough to act — takes a setup period of days to a few weeks, plus an early stretch where you correct it and the corrections stick. Treat the first month as training a very fast, very literal new hire.
This is the right worry, and the answer is the oversight model: start draft-only or approval-per-action, so nothing changes in your systems unseen, and expand permissions routine by routine as it proves accurate. A complete activity log means any error is visible and reversible. Be wary of tools that write to your systems with no log and no approval mode.
The smaller the team, the more the business runs through the owner's memory and phone — which is exactly the load an assistant takes. If your evenings go to invoices, follow-ups, and 'did anyone confirm tomorrow?', size is not the obstacle. The honest threshold is workload shape, not headcount.
Ask directly, per system. Email and calendar are near-universal; mainstream field-service and practice tools are commonly integrated; niche or legacy systems may need browser-based access, which some services support and most self-serve tools do not. A vendor should tell you plainly which of your systems are confirmed, which are possible, and which need discovery.
The good ones adapt to you, not the reverse: if you think out loud in voice notes from the truck, that is the interface; if you live in text, it works there. Any assistant that requires you to learn a new app, new syntax, or new discipline to be useful has moved the admin burden, not removed it.
See it applied
Curious what this would take off your plate?
Skip the sales pitch. See the outcome page for this work, or answer a few questions and get an honest read on where to start.