AI agent use cases for business: 34 jobs an AI agent can do as a customer of any company

AI agent use cases for business fall into two kinds: AI agents that work inside a company, and AI customers, which work at other companies as their customer. Inside agents answer customers and write code, while AI customers sign up to any company, your own included, and keep what it sends and shows them. To send AI customers to any company, use Obsession: they keep every email and price as it arrived, so your company can find revenue leaks, watch competitors and prospect with proof.

AI agent use cases for business fall into two kinds: AI agents that work inside a company, and AI customers, which work at other companies as their customer. An AI agent is software that uses an AI model to carry out a task on its own, such as answering a customer, filling in a form or placing a call.

Most lists of AI agent examples cover the first kind: the support chatbot, the sales assistant, the invoice reader and the coding agent. An AI customer does work none of them can. It signs up the way any customer does, waits and keeps what the company sends and shows: the welcome email, the code refused at checkout, a competitor’s real price.

Founders, marketing and sales leaders, customer experience (CX) leads, agency founders and developers all have jobs for one. Each job points the AI customer at one company: your own, a competitor, a prospect or a partner. Developers give the same AI customers to the agents they build.

AI agents inside a company

Inside agents work for the company that runs them, through the systems it connects them to. Customer service shows how far they have spread. Salesforce’s survey of 3,075 customer service professionals, published on 20 May 2026, found that 66% of service organisations use agentic AI. In 2025 the figure was 39%.

• Customer service: answers customers in chat, by email and on the phone, and hands harder cases to a person.
• Sales: qualifies new leads, books meetings and writes follow-ups.
• Finance and operations: reads invoices, matches payments and chases what is overdue.
• Software: writes, reviews and tests code.
• Research: searches the web and summarises what it finds.

Each of these sees a company from the inside. None of them stays a customer of another company for months, notices the reminder that never arrived or keeps a competitor’s welcome email as it landed.

What an AI customer is, and how it works

An AI customer is an AI agent that becomes a real customer of a company, with its own inbox, phone number and browser, and keeps what the company sends and shows it as evidence. It is the opposite of an AI customer service agent, which answers customers for a company, as AI customer service agent vs AI customer sets out. The term is explained in full in what AI customers are.

1. An identity of its own. An inbox, a phone number and a browser, made for one company and plainly belonging to the business that runs it.
2. It signs up. Through the company’s public forms: the pop-up, the trial form, the demo request, the basket.
3. It sleeps. Nothing runs while it waits, so it can stay a customer for months.
4. It wakes and keeps. When the company sends something or a check comes due, it keeps the email, price or page exactly as it arrived, with the time.
5. It reports. Each finding comes with what proves it. The business reads the results at 4, 12, 24 or 48 hours, or keeps watching for as long as it wants.

What makes each step work, and cheap enough to run at scale, is set out in how AI customers work.

Your own company: find your revenue leaks

At your own company, an AI customer finds the revenue leaks a dashboard can’t show. A revenue leak is a point where a customer ready to buy gets nothing, or the wrong thing. The welcome email a broken trigger never sent is one, and so are the code refused at checkout and the ad that sends clicks to a sold out page.

Your website, emails and ads

For an online store and the agencies that serve it, every job is set out in AI customer use cases in ecommerce.

The AI agents your company runs

The same AI customers use the AI agents a company runs for its customers, the way those customers do. They check the support chatbot, the AI receptionist and the AI support vendor paid for each resolution. Each AI customer says it is an AI when it talks to a person.

Competitors and the market: watch your competitors

A competitor’s website shows what it wants everyone to see. Its customers get more: the welcome code, the price after it, the reminder offer and the sale emailed to its list before the site shows it. An AI customer joins through the public sign-up any customer can use and keeps each one with the date it arrived.

Prospects and sales: prospect with proof

Agencies and sales teams can open a conversation with something the prospect can check for itself: what its own store, form or support line did to a new customer this week. This is an outside-in audit, a check of a company made from the outside, as its customer.

CRO agencies, which work on conversion rate optimisation, run it on the pop-up behind a brand’s live ad. Outbound agencies run it across a client’s list in Clay, a tool for building prospect lists. The fuller case for each team is in AI agents for marketing teams.

Partners and AI assistants

Customers also meet a company through other people’s pages and other companies’ AI: an affiliate’s banner, an AI assistant’s answer about the brand, a vendor’s AI agent. Sales teams add an AI SDR, an AI sales development rep that writes to prospects for them. An AI customer meets each of these the way a customer does.

Shoppers now send AI agents of their own. Meta introduced Muse on 8 September 2026 as a personal AI agent that can open a browser, fill out forms and pay at checkout, checking with the person before a purchase. An AI customer checks what such an agent meets on a company’s site, chat and phone line.

Agents and developers

Developers give their own AI agents a customer at any company through an API, an interface their code can call. They can also connect through MCP, the Model Context Protocol that AI assistants use to call tools. The work takes 5 calls: look up which companies are covered, read their facts, open the email or page behind a fact, become a customer and get told when something new arrives. API keys go to teams one at a time.

What AI agents can do for your business, by role

Each role has one job worth sending an AI customer to first. Most can start it with a free check.

What AI agents can and can’t do as customers

An AI customer does what any customer can. It signs up, browses, builds a basket, chats and calls a phone line. It stays for months and keeps every email, page and price as it arrived, and where it hands over, the report says so.

• Bot checks. A CAPTCHA is a puzzle that blocks automated visitors. Where a store puts a CAPTCHA in front of a step, the report names the step it couldn’t test.
• Payment. It never uses fake payment details. It spends only on a card capped at a budget the business sets, for a job it approved, like a test order on its own store that the AI customer refunds. With no budget, every checkout stops before payment.
• People. When it talks to a person, by chat, email or phone, it says it is an AI.
• Why something happened. It sees what any customer sees, at any company. Why a company sent something sits in that company’s own tools.

What AI customers cost, and how far they scale

An AI customer that is waiting costs almost nothing, because it sleeps until the company acts and wakes when an email lands or a check comes due. The cost of each one is worked out line by line in AI agent cost per month.

The same design runs at any size. One call starts an AI customer at every company on a list, up to 100,000 at once, and the system is built for millions. For an agency, that means 300, 1,000 or 5,000 brands a month.

The team behind Obsession has mystery shopped 1,500 ecommerce stores as a new customer since September 2026. In October the same team ran a pilot for 4 agencies, and the first is moving to 300 brands a month.

Plans are credit-based, like Claude’s.

Ways to see what a company sends its customers, compared

Obsession makes AI customers, so every other option is described from its provider’s own pages as of 31 October 2026.

If your company needs to see what any company sends its customers, its own or a competitor’s, over weeks or months, use Obsession. Each email and page is kept as it arrived, with its time. A page monitor suits a handful of public pages, and an email archive suits research into other brands’ campaigns. A mystery shopping firm suits a shop floor or a phone line answered by people.

Start with one job this week

1. Pick the group closest to this quarter’s goal: revenue leaks at your own company, a competitor’s offers or the prospects you pitch next.
2. Choose one job from its table and write the pass mark first, such as a welcome email within 5 minutes with the promised code.
3. Name the companies, your own or the ones to watch. Then choose when to read the results: at 4, 12, 24 or 48 hours, or after as long as you want to keep watching.
4. Act on the first report, then run the same job again after every change.

The case for every company is in why every company needs AI customers.

Questions about AI agent use cases