If you want AI to fix a task in your business, the honest answer depends on what the task really is. AI genuinely helps when the work starts with messy information, such as PDFs, photos and emails, that a person has to read. But many tasks that feel like an AI problem are really something else: a simple rule-based automation, two systems that need connecting, data that needs cleaning, or a report nobody has built yet. Those are often quicker and cheaper to fix, and fixing them first usually makes any AI you add later work better. You can find out which one you have in two minutes with our AI fit check.
Why "we need AI" is a good place to start
Wanting AI usually means something real: the team is busy, the same work comes round every week, and someone has noticed that a lot of it is copying, checking and chasing. That instinct is right. The question is which tool fits the work.
The five situations below cover most of what we hear from growing businesses with an office team, several systems (accounts, jobs, stock, a CRM) and a few spreadsheets holding it together. Each one has a different fix.
1. AI genuinely helps: messy information that a person has to read
What it looks like: supplier invoices arrive as PDFs in a dozen different layouts. Drivers send photos of signed delivery dockets. Customers email purchase orders. Someone reads each one and types the details into another system.
How to tell: the information comes from outside your business, every sender lays it out differently, and the hard part is reading and understanding it. A person could check the result quickly.
What fixes it: AI that reads the document, pulls out the details and prepares an entry for a person to approve. This is a real, practical use of AI today, and it's where most businesses should start. See AI document processing for invoices, dockets and forms, AI workflow automation for wider uses, or the back-office AI pilot to try it on one task with a person checking every result.
Other good fits: sorting incoming emails and requests, matching receipts to jobs, and drafting a monthly summary for someone to review.
2. A simple rule or automation does it
What it looks like: every approved quote has to become a job, every new job needs a folder and a checklist, and invoices over a set amount go to the owner for approval. Someone does each step by hand.
How to tell: the task follows the same steps and rules every time. You could write it down as "if this, then that" without any "it depends".
What fixes it: a rule-based automation. It doesn't need AI, it does the same thing every time, it costs less to run and it's easier to check. Using AI here adds cost and a small chance of a wrong answer for no benefit. See business process automation.
3. Your systems need connecting
What it looks like: a job is entered in the job system, then retyped into the accounting system to invoice it. Stock levels are copied from one system to another at the end of each day. Someone exports a spreadsheet from one system and imports it into another.
How to tell: the information already exists in one of your systems. The work is carrying it into another one by hand. This is often called double entry (the same information typed into two systems).
What fixes it: connecting the systems so information is entered once and flows to where it's needed. AI would only be a more expensive way of retyping it. See systems integration.
4. Your data needs cleaning first
What it looks like: the same customer appears three times with slightly different names. Item codes don't match between the stock system and the accounts. Reports from two systems give different totals, and nobody is sure which one is right.
How to tell: people don't trust the numbers, spend time checking them against each other, or keep their own spreadsheet "just in case".
What fixes it: cleaning the data before doing anything else: removing duplicates, agreeing who owns each list, and putting simple rules in place so it stays clean. AI and automation both work on the data you give them; if it's wrong, they just process the mistakes faster. See data cleansing services.
5. You need a report, not AI
What it looks like: the owner wants to know which jobs are making money, how much stock is tied up, or what cash will look like next month. Someone spends a day each month building a spreadsheet to answer it, and by the time it's ready it's out of date.
How to tell: the real problem is seeing the numbers, not doing a task. People ask for "AI insights" when what they need is one report they can trust.
What fixes it: a report that pulls the figures from your systems and updates itself, so everyone works from a single source of truth (one place everyone takes the number from). See one set of numbers everyone trusts.
Why fixing systems and data often comes before AI
These five situations often turn up together. For example, a business might want AI to read delivery dockets (situation 1), but the details then have to be retyped into the accounts (situation 3), and the customer list has duplicates (situation 4). AI can still read the dockets, but most of the time saved is lost if a person then retypes the result into a system with messy data.
That's why the order matters. Connect the systems and clean the data that the task depends on, and AI then has somewhere reliable to put its results. The AI part becomes smaller, cheaper and easier to trust.
It doesn't mean AI has to wait for everything else to be perfect. Reading documents can start on its own, with a person checking the results. It just means looking at the whole task before choosing the tool.
How to work out which one you have
Pick one task your team does every week and answer these questions:
- Where does the information come from: documents and emails, a form, another system, or a spreadsheet?
- Does it happen the same way every time?
- Is the same information typed into more than one system?
- Is the real problem seeing the numbers?
- How much judgement does it need?
- Could a person check the result quickly?
Our free AI fit check asks exactly these questions and gives you a plain answer, with the reasons and a sensible first step. It takes about two minutes and nothing is sent to us.
Frequently asked questions
Is automation the same as AI?
No. Automation follows rules you set: when this happens, do that. AI handles work that needs reading or interpreting, such as working out what an email is asking for or which number on an invoice is the total. Many good systems use both: AI reads the document, then a rule-based automation sends it to the right person and into the right system.
Should we start with AI or with integration?
Start with whatever the task actually needs. If the information comes in as documents that someone has to read, AI is a sensible first step. If it already sits in one of your systems, connect the systems first. When both apply, connecting the systems usually saves more time.
Do we need clean data before using AI?
For anything that puts results into your systems, it helps a lot. AI that reads supplier invoices still has to match each one to a supplier, and if that supplier is in your system three times, someone has to sort it out. Cleaning the lists the task depends on is usually a small job compared with the time it saves.
Can AI build our reports for us?
AI can draft a summary of figures for someone to review, but it can't fix numbers that disagree between systems. A reliable report comes from connecting the systems and agreeing where each number comes from. Once that's in place, AI can help explain what changed.
What does it cost to find out?
Nothing to start. The AI fit check is free, and so is an initial chat. If you want it worked out properly across all your team's tasks, our AI readiness assessment is delivered through the fixed-price $5,000 Process & Systems Audit.
If you'd like a second opinion on a task, book a free initial chat. Tell us about the task and the systems involved, and we'll give you an honest view of what would fix it, whether that's AI or not.