How to use AI in a small business

This article breaks down how small business owners can get maximum value from AI without wasting time. It includes ready-to-use prompts, as well as quick frameworks for setting up an AI pilot project, calculating which projects are worth automating, and deciding when to use AI or hire someone.

What small businesses can use AI for

The sections below cover a small business's most time-consuming functions. Each one names what AI actually helps with, an example query, and where a person still needs to step in.

Customer support

Support inboxes are full of repetitive, time-sensitive requests, which makes this one of the easier places to start.

AI tools can help you:

• Draft responses to routine questions
• Summarize customer histories
• Classify and route incoming requests
• Search a knowledge base for relevant information
• Identify messages that need urgent human attention

Try this query:

Review the customer-support emails and attached FAQs. Group the inquiries into recurring topics, identify questions the FAQ does not answer and create a spreadsheet with the customer name, issue, urgency and recommended next action. Draft response options for routine questions, but do not send anything or make refund decisions. Flag complaints, policy exceptions and legal threats for human review.

Don’t use AI to:

• Resolve disputes
• Make refund or compensation decisions without authorization
• Respond to emotionally sensitive complaints without review
• Replace a clear route to a human representative

Marketing

Marketing is usually the first place small businesses turn to AI. There’s far more to it than creating email copy, however.

AI tools can help you:

• Write a first draft of a website page or landing page from a plain description of the business
• Map out which pages a website needs, like a separate page for each service, product, or location
• Spot gaps in the website: missing details, outdated pages, or topics customers ask about that aren't covered anywhere
• Turn common customer questions into a list of blog posts or pages worth writing, prioritized by what matters most to the business
• Pull together a report on how the website is doing, like which pages get traffic and how easily people can find the business in a search

Try this query:

"Review this business information, existing website and list of services. Research the most important customer questions and competing pages for '[topic or service].' Create a recommended site structure with page titles, target intent, primary and supporting topics, calls to action and internal-link suggestions. Then draft the homepage and one service page using the approved brand guidelines. Flag every claim that needs confirmation, and do not publish changes."

Don't use AI to:

• Build a differentiated brand voice from scratch
• Replace customer research or strategic positioning
• Publish without fact-checking and editorial review
• Produce large volumes of generic content solely for search traffic

Scheduling and administration

Calendars, inboxes, and meeting notes eat hours that don't show up on any invoice, which makes admin a strong use case for AI.

AI tools can help you:

• Summarize meetings and extract action items
• Draft follow-up emails
• Organize documents and forms
• Identify calendar conflicts and scheduling gaps
• Classify incoming requests
• Connect customer inquiries to internal tasks

Try this query:

"Review this week's calendar, project notes and open email threads. Create a prioritized list of commitments, overdue actions, scheduling conflicts and unanswered questions. Prepare a one-page weekly operations report with links to the relevant source documents. Draft, but do not send, follow-up emails for the three most urgent items."

Don't use AI to:

• Make priority calls without business context
• Commit the business to deadlines or prices without approval
• Handle sensitive employee or customer information in an unsuitable tool
• Treat a summarized meeting record as complete or accurate

Finance and bookkeeping

Bookkeeping tasks are repetitive and rule-based, which is exactly the kind of work AI handles well, with an accountant checking the output.

AI tools can help you:

• Categorize transactions for review
• Identify overspending
• Flag unusual spending or duplicate entries
• Summarize cash-flow patterns
• Extract data from receipts and invoices
• Compare sales or expenses across periods
• Prepare questions for an accountant or financial adviser

Try this query:

"Analyze the attached transaction export for the last six months. Flag unusual changes, possible duplicate transactions, missing categories and expenses that require accountant review. Create a spreadsheet with the flagged items and a short summary of the main patterns. Do not make tax recommendations, recategorize transactions automatically or send the report externally."

Don't use AI to:

• File taxes
• Make final accounting judgments
• Decide whether the business can afford a major expense
• Provide legal, tax, or financial advice without qualified review

Sales

Leads can go cold fast, and AI is good at catching the ones a busy owner doesn't have time to chase.

AI tools can help you:

• Spot leads you haven't followed up with in a while
• Summarize a sales call so you don't have to relisten to it
• Draft a personalized follow-up email for each lead, instead of a generic one
• Spot the reasons prospects keep giving for not buying, so you can address them directly
• Flag deals that have stalled and need a nudge
• Turn your notes into a first draft of a proposal

Try this query:

"Review the attached sales notes and identify opportunities with no follow-up in the last 14 days. Research each company using current public sources and create an account brief with its business priorities, relevant developments, likely pain points and evidence links. Draft a personalized follow-up email for each opportunity using only verified information. Do not send the emails or make claims about budget, authority or purchase intent."

Don't use AI to:

• Negotiate pricing or contract terms without authorization
• Make promises about delivery or product capabilities
• Close complex deals independently
• Replace relationship-building in consultative sales

Hiring and onboarding

Hiring paperwork and onboarding materials follow a template most of the time, which leaves room for AI to draft the first pass.

AI tools can help you:

• Draft job descriptions
• Create interview-question templates
• Turn existing policies into first-draft onboarding materials
• Summarize training content
• Build role-specific checklists

Try this query:

"Using the approved role profile, company handbook and onboarding checklist, create a first-week onboarding plan for a new operations coordinator. Draft a role overview, training schedule, checklist and manager briefing. Identify any conflicting or outdated information in the source documents. Do not screen applicants, rank candidates or make hiring recommendations."

Don't use AI to:

• Screen candidates without human oversight
• Make hiring decisions
• Infer protected characteristics or personal traits
• Replace reference checks or structured evaluation

Which small-business AI tools do you need?

AI tools go by different names depending on what they do: some are general-purpose assistants, others are built into software you already use, and a growing category runs multi-step tasks on its own. The table below breaks down which type fits each job, and what to check before you commit to one.

How to start using AI in your small business

The first question is not which AI tool to buy. It is which workflow is creating a measurable bottleneck for the business. With that information, put together a small-scale pilot project:

1. Map the bottleneck

List out where time is actually being lost, like an inbox that takes an hour to clear every morning, invoices that sit unprocessed for a week, or follow-ups that fall through the cracks.

2. Choose one workflow

From that list, pick the task that's most consistent and lowest-risk to start with, not necessarily the one costing the most time. A repetitive, well-defined task is easier to test than one that needs judgment every time.

3. Set a baseline

Record how long the task currently takes, what it costs, and how often it goes wrong, so there's something to compare the AI version against.

4. Check the data

Confirm the information the tool will use is accurate and up to date, and that it only has access to what it actually needs, not your entire inbox, drive, or customer database by default.

5. Select a tool that fits the existing tech stack

Favor tools that connect directly to the calendar, inbox, CRM, or accounting software already in use, rather than ones that require exporting and importing data by hand.

6. Assign a reviewer

Decide who checks the output before it goes out, and who handles the cases the tool gets wrong.

7. Run a limited test

Pick a small group of users, a fixed time period, and a clear measure of success before rolling it out to the rest of the team.

8. Measure and adjust

Compare the results against the baseline, then keep the workflow, revise it, or drop it based on what the data actually shows.

When should you use AI in your small business?

AI is worth testing on tasks where checking the AI's work takes less time than doing the task by hand. A few signs point to a good fit:

• The task happens several times per week
• The process is reasonably consistent
• The output can be checked quickly
• Occasional errors are recoverable
• The business can measure improvement
• The tool fits existing systems and does not require excessive setup

A task is a poor fit for automation when:

• It’s done infrequently
• It requires nuanced judgment or emotional intelligence
• An error would create legal, financial, or reputational damage
• The work is public-facing and difficult to correct
• The process is still changing rapidly
• The data is incomplete or unreliable

Analyzing the ROI of using AI

Many individual AI tools begin in the low tens of dollars per month. But, when calculating the cost-benefit of automating a task, consider that the main cost may be the setup time, integration, training, review, and correction involved.

Here's a simple way to run the math: figure out the payback period, or how long until the time saved outweighs the time spent setting up.

• Estimate the weekly cost of doing the task by hand: how long one instance takes, times how often it happens in a week, times what that person's time is worth per hour.
• Estimate the weekly savings from AI: the manual cost, minus the cost of reviewing the AI's output each time, times how often it happens.
• Divide the one-time setup and integration cost by that weekly savings. That's roughly how long it takes to break even.

Here’s an example, assuming the person who normally does the task is paid $30/hour:

A task that takes 20 minutes and happens twice a week costs about $20 a week done by hand. If reviewing the AI's output still takes 10 minutes each time, that only saves about $10 a week.

Two hours of setup (worth $60 at that rate) would take six weeks to break even, too long for a first pilot.

A task that takes an hour and happens daily costs about $150 a week done by hand. If reviewing the AI's output takes 15 minutes each time, that saves close to $120 a week. A few hours of setup pays for itself in under a week.

When should you hire instead of automate?

Hire when the work requires ongoing judgment, relationship-building, accountability, or increasing capacity. Use AI first when the task is well-defined, repetitive and not yet large enough to justify a dedicated role.

Hiring and AI are not always substitutes for each other. A new employee who uses AI can handle more work, document processes, analyze information, and improve service faster than one working without it. The question is not always “Should we hire or use AI?” It may be “What should a person own, and which parts of that role can AI support?”

Putting this into practice

Pick the task that's costing the most time and follows a repeatable process, run the payback math, and pilot it with a reviewer in place before it touches a customer or a dollar.

Perplexity Computer gives a small team more capacity without adding headcount. It connects to more than 400 tools small businesses already run on, including QuickBooks, Mailchimp, Shopify, and Stripe. From there, it pulls the relevant data, runs the analysis, and drafts the output, flagging anything that needs a signoff before it goes out.

Try Perplexity Computer for your small business.

FAQs

Is AI worth it for a small business? It depends on the task. AI is worth testing when the task happens often, follows a clear process, and takes less time to review than to do by hand. It is a weaker fit for rare, high-judgment, or high-stakes work.

What is the best AI tool for a small business? There is no single best tool. The right one depends on the workflow: help-desk AI for support inboxes, workspace assistants for scheduling and email, accounting software with AI features for bookkeeping, and agentic platforms for tasks that span several tools at once.

How much does AI cost for a small business? Many tools start in the low tens of dollars a month, but the full cost includes setup, integration, training, and review. A low subscription price does not guarantee a positive return if the tool needs heavy correction.

Can AI replace employees at a small business? AI is better suited to supporting a role than replacing one, particularly where judgment, accountability, or relationship-building matter. A new hire who uses AI can often handle more work than either a person or a tool alone.

When should a small business avoid automating a task? Skip automation when the task is rare, requires nuanced judgment, would be costly to get wrong, is public-facing and hard to correct, or depends on data that is incomplete or unreliable.