Most business owners I speak to already know AI could help them. What they don't know is where. And without a clear answer to that question, nothing moves. You end up reading about AI, attending webinars about AI, maybe playing with ChatGPT for a week, and still running the same manual processes you were running two years ago.
An AI audit cuts through that. It's how you go from "AI is probably useful for us" to "here are the three processes we should automate first, and here's what it's costing us not to."
At AMPL, we run an operational audit with every client before we build anything. Not because it's a nice-to-have, but because without it you're guessing. This post walks through what that audit actually looks like, what it typically uncovers, and how you can run a basic version yourself.
What an AI audit actually is (and what it isn't)
An AI audit for business is a structured review of your operations to find where AI can meaningfully reduce manual effort, cost, or error, and where it can't. It looks at your processes, your data, your tools, and how your team spends their time.
It's not a technology assessment. It's not about which AI tools are trending or whether your software is "AI-ready" in some abstract sense. And it's definitely not a pitch document for a particular platform.
A good audit is diagnostic. It finds the drag in your operations. Some of that drag is automatable. Some of it is better handled by augmenting what your team already does. Some of it just needs a cleaner process. The audit tells you which is which.
AI audit vs AI strategy vs AI implementation
These three things get conflated a lot, so it's worth being clear about what each one is.
An AI audit is the diagnostic phase. It answers: where are the opportunities, and how big are they?
An AI strategy takes the audit findings and builds a plan. It prioritises which opportunities to pursue, in what order, and at what cost. It's the bridge between knowing and doing.
An AI implementation is the actual build, the systems, automations, and agents that deliver the outcome.
You can't do strategy without the audit. And implementation without strategy is how you end up with expensive tools that don't actually solve anything. The audit is where it has to start.
The four areas every AI audit should examine
When we run an audit with a new client, we look at four things consistently. These are the areas where we reliably find waste, bottlenecks, and automation opportunity.
Process inventory — finding the manual work hiding in plain sight
The first job is just listing what actually happens in the business each week. Not the high-level stuff from a company overview, the actual granular tasks that someone does.
What we usually find is that a significant chunk of those tasks are repetitive, rules-based, and deeply manual. Things like copying information from an email into a CRM, chasing clients for documents, reformatting data between systems, generating the same report every Friday morning.
These tasks are invisible in most businesses because they've become habits. Nobody questions them. They just get done. The process inventory makes them visible.
With a removals company we worked with, the process inventory alone surfaced about 14 hours a week of manual admin across their small team, most of it in booking, job scheduling, and client communication. None of it was being tracked. It was just absorbed into the working day.
Data availability — do you have what AI needs to work
AI systems need data to operate. Before you can automate a process, you need to understand what data that process relies on, where it lives, and whether it's in a usable state.
This catches a lot of businesses out. They want to automate their quoting process, for example, but their historical quote data is in a mix of PDFs, spreadsheets, and someone's memory. That's not a dead end, but it does affect the timeline and approach.
The data availability check asks: is the information structured or unstructured? Is it centralised or scattered? Is there enough volume to learn from? The answers shape what's possible and how quickly.
Tool sprawl — where your stack is creating bottlenecks
Most growing businesses end up with a collection of tools that were each adopted to solve a specific problem, and that now don't talk to each other. You've got a CRM over here, a project management tool over there, invoicing in one place, files in another. And staff are acting as the connective tissue between them, manually moving information from one system to the next.
Tool sprawl is one of the most consistent findings in any AI readiness assessment. And it's one of the clearest automation opportunities, because integration is something AI and automation do well.
We look at what tools are being used, what data flows between them, and where the manual handoffs are happening. Usually three or four of those handoffs are candidates for immediate automation.
Staff time mapping — where hours are going each week
This is the most direct way to quantify opportunity. If you can map where your team's time is actually going, not where it's supposed to go, where it actually goes, you can put a number on the cost of your current processes.
We ask teams to track their time for a week, or in some cases we reconstruct it from calendar data and conversations. The pattern that comes out is almost always the same: a disproportionate amount of time in a handful of repetitive tasks that people have stopped questioning.
When we worked with a financial services business on their client onboarding process, the time mapping showed their team was spending close to 40% of their week on tasks that were either directly automatable or could be dramatically shortened with the right tooling. That's not unusual. It's actually fairly typical.
How to run a basic AI audit yourself (step-by-step)
You don't need a consultant to run a basic version of this. Here's a process you can work through in a few hours, ideally with one or two of your team involved.
Step 1 — Map your weekly operations
Start by listing every recurring task your business does in a typical week. Go function by function: sales, operations, finance, client communication, HR, reporting. Don't filter anything out at this stage. The goal is completeness, not prioritisation.
For each task, note roughly how often it happens, who does it, and approximately how long it takes. A simple spreadsheet works fine. You're building a process inventory, not a formal process map.
Step 2 — Tag every task as automate, augment, or keep human
Go through your list and assign one of three tags to each task.
Automate: The task is repetitive, rules-based, and doesn't require judgement. Data entry, report generation, email routing, document formatting, these are strong automate candidates.
Augment: The task requires human judgement, but AI can make it faster or better. Writing client proposals, reviewing contracts, answering complex queries, AI can assist here without replacing the human entirely.
Keep human: The task is relationship-based, genuinely creative, or requires contextual judgement that AI can't reliably replicate. Senior client relationships, strategic decisions, nuanced negotiations, keep these human.
Most businesses find that 20-35% of their weekly tasks fall into the automate category once they look honestly.
Step 3 — Score by effort vs return
Not every automation opportunity is worth pursuing first. Take your automate and augment list and score each item on two axes: how much time or cost it would save (return), and how complex it would be to automate (effort).
High return, low effort tasks are your quick wins. Start there. High return, high effort tasks go on the roadmap for later. Low return tasks, regardless of effort, deprioritise or ignore.
What you end up with is a prioritised shortlist of two or three processes to actually act on. That's more valuable than a long list of theoretical possibilities.
What an AI audit typically uncovers
After running audits across businesses in insurance, construction, removals, financial services, and professional services, the findings tend to cluster around the same types of waste.
Email triage and routing. Someone, usually an ops person or the owner, is spending significant time reading, sorting, and forwarding emails. This is almost always automatable.
Quote and proposal generation. Businesses that produce custom quotes are often rebuilding them from scratch each time. The inputs change, but the structure doesn't. Strong candidate for AI-assisted generation.
Data entry between tools. Information from one system that needs to appear in another, manually copied by a human. Every time we see this pattern, it's automatable.
Manual reporting. Someone pulling data from multiple sources, formatting it, and sending it somewhere on a regular cadence. This is almost entirely automatable.
Client onboarding and document collection. Chasing clients for information, tracking what's been received, following up. This whole workflow can be systematised and automated.
With one client in the equipment supply space, the audit surfaced that their team was spending roughly 60% of their operational time on tasks that fell into these five categories. The build we delivered recovered about 25 hours a week across their team, roughly £40,000 a year at their staff cost rates.
That's not an unusual outcome. It's what you find when you actually look.
When to do it yourself vs hire someone
The DIY version above is genuinely useful. If you're not sure whether AI is relevant to your business, or you want to build the internal case before spending anything, run through those three steps yourself. You'll come out with a clearer picture than 90% of businesses have.
But there are situations where an external audit is worth it.
If your operations are complex, multiple teams, multiple systems, significant process interdependencies, the DIY version will miss things. Not because the framework is wrong, but because you're too close to your own operations to see all of it clearly. An outside perspective catches things that internal reviews don't.
If you've already tried to map this and got overwhelmed, or if you ran the analysis and ended up with a long list but no clear priority, that's also a sign that a structured external audit will save you time.
And if you're planning to invest significantly in AI or automation, whether that's a custom build, a new platform, or a team hire, it's worth doing the audit properly first. The cost of an audit is small relative to the cost of building the wrong thing.
At AMPL, our audit goes through your operations in detail, maps the opportunities, scores them by ROI, and gives you a specific costed set of recommendations. It's refundable against your first build if you decide to proceed, so there's no risk in finding out.
FAQ
What is an AI audit for business?
An AI audit for business is a structured review of your operations to identify where AI and automation can meaningfully reduce manual work, time, or cost. It examines your processes, data, tools, and staff time to produce a prioritised list of automation opportunities, specific to your business, not a generic checklist.
How long does an AI readiness assessment take?
A basic self-directed assessment can be done in a few hours using the process outlined above. A thorough external audit, the kind that covers complex operations in detail, typically takes one to two weeks, including discovery calls, process mapping, and the recommendations writeup.
What does a business process audit for AI actually look at?
It looks at four things: your process inventory (what tasks happen each week), your data availability (what information AI would need and whether you have it), your tool stack (where systems aren't talking to each other), and your staff time map (where hours are actually going). Together, these surface the real automation opportunities.
How do I know where to use AI in my business?
Start by listing your recurring weekly tasks and tagging each one as automate, augment, or keep human. Automate candidates are repetitive, rules-based tasks that don't require judgement. Augment candidates need human input but could be faster or better with AI assistance. Then score your list by effort vs return to find where to start.
What does an AI opportunity assessment typically uncover?
The most common findings are email triage and routing, manual data entry between tools, quote or report generation, and client onboarding workflows. These patterns appear consistently across industries. Most businesses find that 20-35% of their weekly tasks are candidates for automation once they look carefully.
Should I hire someone to run an AI audit or do it myself?
If your operations are relatively straightforward, the DIY version in this post will give you a solid starting point. If you have multiple teams, complex interdependencies, or you're planning a significant investment in AI, an external audit is worth it. You're too close to your own operations to catch everything, and the cost of building the wrong thing is much higher than the cost of the audit.
