We ended our recent post on what MCP is and why it matters with a line that a lot of people picked up on: the businesses that'll benefit most from AI are the ones that get their data in order now. Which is good advice, as far as it goes. But a few people rightly came back and asked, "Right, but how exactly do we do that?"
Fair point. It's easy to say "sort your data out" and leave everyone nodding vaguely while nothing actually changes. So this is the companion piece. The practical, roll-your-sleeves-up guide to getting your business into a state where AI tools can actually do something useful for you, rather than just generating impressive demos that fall apart the moment they meet your real systems.
We've spent the last while connecting AI to real business systems (our own included), and we've learned first-hand what makes the difference between AI that works and AI that's just an expensive novelty. Most of it has nothing to do with the AI itself. It's the boring stuff underneath.
Why AI readiness is mostly about you, not the AI
Here's something that doesn't get said nearly enough: AI tools are already good enough for most business use cases. The technology isn't the bottleneck. You are. Or more specifically, your data is, your processes are, and the way your systems talk to each other (or don't) is.
There's a stat that gets thrown around a lot in the tech world, that something like 60% of AI projects get abandoned before they deliver any value. And the reason is almost never that the AI wasn't clever enough. It's that the business wasn't ready. The data was a mess, nobody could agree on what success looked like, or the systems that needed to connect were so siloed that no AI tool could reach the information it needed.
So before you spend a penny on AI tools, subscriptions, or consultants, here's what to actually do. Think of it as laying the foundations before you build the house.
Step 1: Work out where you're wasting the most time
This sounds obvious, but almost nobody does it properly. Before you think about what AI could do for you, you need to understand where your people are spending their time on things that shouldn't need a human.
Sit down with your team leads and ask one question: "What's the most tedious, repetitive part of your job?" You'll hear things like manually copying data between systems, chasing people for updates, reformatting reports, and pulling together information from five different places to answer a simple question. Write it all down. Every single one.
Then rank them. Not by what sounds most exciting to fix, but by how much time they actually cost. If your accounts team spends three hours every week manually cross-referencing invoices with project records, that's 150 hours a year. That's nearly a month of someone's working life. That goes near the top of the list.
This list becomes your AI business case. It's also the thing that stops you from getting seduced by flashy AI features that don't actually solve a problem you have. We've seen plenty of businesses get excited about AI-powered content generation, only to find that the real time sink is their quoting process. Solve the expensive problem first.
Step 2: Audit your systems (all of them)
Now you need to know what you're working with. Make a list of every system your business uses. CRM, accounting software, project management, email marketing, your website CMS, any spreadsheets that people rely on, shared drives, even that database someone built in Access ten years ago that everyone pretends doesn't exist. Especially that one.
For each system, write down four things:
- What data does it hold?
- Who puts data into it (and how often)?
- Who gets data out of it (and how)?
- Does it connect to anything else, and if so, how?
If you're using something like HubSpot as your CRM, you're already in reasonable shape on the integration front, because platforms like that are built to connect to other tools. But you'd be amazed at how many businesses have a perfectly good CRM that half the team doesn't use because nobody enforced the habit, so the real customer data lives in Outlook inboxes and handwritten notes.
The goal of this audit isn't to judge anyone. It's to get an honest picture of where your data actually lives, versus where you think it lives.
Step 3: Clean up your data (yes, this is the boring bit)
We're not going to pretend this is fun. But it is, without question, the single most important thing you can do to make AI useful for your business.
AI tools learn from and work with your data. If your data is inconsistent, incomplete, or scattered across systems that don't talk to each other, the AI's output will reflect that. Garbage in, garbage out, as they used to say in the early days of computing. Turns out that's still true, even with very clever AI.
Here's what "cleaning up your data" actually means in practice:
Start with how things are entered. If your CRM has "Aberdeen" and "aberdeen" and "Abdn" and "AB10" all meaning the same place, that's a problem. Pick a format and stick to it. The same goes for company names, job titles, product codes, and anything that someone might enter slightly differently each time.
Then look at the gaps. Check your key records (customers, projects, products) and see what's missing. If half your customer records don't have an industry field filled in, or your project records don't consistently track which service was delivered, fix that before you try to get AI to analyse anything.
Get rid of the dead wood while you're at it. Duplicate records, contacts who left their company five years ago, project records from systems you don't use any more. Clean them out. The less noise in your data, the more useful anything built on top of it will be.
And make sure the same words mean the same things everywhere. If "revenue" means one thing in your accounting system and something slightly different in your sales pipeline, you've got a problem that no AI tool will magically resolve. Get your definitions agreed across the business before you start connecting things.
This is unglamorous work, and it takes time. But we've seen businesses try to skip this step and then wonder why their shiny new AI tool keeps giving them rubbish answers. Every single time, the answer is the data.
Step 4: Map how your systems connect (or don't)
This is where it gets interesting, because this is the step that directly determines what AI can and can't do for you.
Take your list from Step 2 and draw out how data flows between those systems. Does information from your website forms automatically end up in your CRM, or does someone copy it over? When a deal closes in your sales pipeline, does the project management system know about it, or does someone create a new project from scratch? When you raise an invoice, is the finance system pulling from the same data that the project team used?
Every manual handoff, every copy-and-paste, every "oh, we export a CSV and then import it into the other system" is a point where data gets lost, mangled, or delayed. These are also not coincidentally exactly the points where AI can make the biggest difference, but only if you understand where they are.
If you read our MCP explainer, you'll remember that MCP gives AI tools a standard way to talk to your business systems. But the AI can only talk to systems that are willing to converse. If your key data is locked in a spreadsheet on someone's desktop, no protocol in the world will help. The prep work you do now, getting data into proper systems with proper connections, is what makes that future possible.
Step 5: Pick one thing and start small
This is where we most often see businesses go wrong. They do all the groundwork, get excited, and then try to "implement AI across the organisation" all at once. That's a recipe for expensive disappointment.
Go back to your list from Step 1. Pick the biggest time-waster that also has reasonably clean data behind it (from Steps 2 and 3). That's your pilot project.
For a lot of the businesses we work with, the first win looks something like this: connecting their CRM to their AI assistant so that it can pull together client summaries, draft responses, or flag overdue follow-ups. It's not glamorous, but it's immediately useful, and it teaches the whole organisation what "AI that actually works" looks like.
Keep the scope tight. Measure the time saved. Get the team's honest feedback (not just from the people who were excited about it, but from the ones who were sceptical too). Then, once it's working, expand from there.
If you're in B2B and wondering what this looks like from a marketing angle, our AI search strategy guide for marketers covers how AI is changing the way businesses find and engage with their audience. It's a useful companion read if your first AI project ends up being marketing-related.
Step 6: Get your people on board (properly)
We've deliberately left this until Step 6 because, in our experience, you need something concrete to show people before this conversation goes anywhere useful. If you stand up in front of your team and say "we're adopting AI" before you've done any of the groundwork, you'll get a mixture of eye-rolling, panic, and enthusiastic suggestions for AI tools that solve problems you don't have.
Once you've got a pilot running (or at least mapped out), have honest conversations with your teams about what AI is and isn't going to do. It isn't going to replace anyone's job (and if someone tells you otherwise, be very sceptical). What it is going to do is take the most tedious parts of people's roles and handle those, so they can spend their time on the work that actually needs a human brain, the judgment calls, the relationship-building, the creative thinking.
The businesses that get the most from AI are the ones where the team actively looks for ways to use it, where they say, "Could the AI handle this bit?" in everyday conversations. That only happens if people aren't afraid of it and understand what it's for. So invest time in explaining, demonstrating, and listening to concerns. Don't just announce it and hope for the best.
Step 7: Think about what "good" looks like
Before you get too far down the road, agree on how you'll know whether this is working. And make it specific. "We want AI to make us more efficient" is not a goal, it's a wish. "We want to reduce the time spent on monthly client reporting from 15 hours to 3 hours" is a goal. You can measure it, track it, and know when you've hit it.
Set these targets before you start, not after. It's surprisingly easy to convince yourself that an AI project was worthwhile after you've spent the money, even when the numbers don't really back it up. Decide in advance what success looks like, and hold yourself to it.
What you can do on Monday morning
If you've read this far and you're wondering where to start, here's your to-do list for the coming week:
- Ask your team leads: "What do you spend the most time on that a computer should be able to do?" Write down every answer.
- List every system your business uses. Every single one, including the spreadsheets and the legacy tools that everyone's embarrassed about.
- Pick the three systems where your most important data lives and check whether they can connect to each other. If they can't, that's your first problem to solve.
- Look at the data in your CRM. Is it complete? Is it consistent? Is it up to date? If the answer to any of those is "not really", block out some time to fix it.
None of that requires buying anything or hiring anyone. It's just looking clearly at what you've got and being honest about the state it's in. And that honesty is genuinely the hardest part, because most of us have been papering over data problems for years.
If you're curious about how AI agents (the tools that actually do things in your systems, rather than just chatting) fit into all this, our plain English guide to AI agents explains the difference and why it matters.
The bottom line
AI readiness isn't a technology project. It's a housekeeping project. The businesses that'll get the most from AI over the next few years aren't the ones with the biggest IT budgets or the most advanced tools. They're the ones that took the time to sort out their data, understand their systems, fix the processes that were already a bit broken, and get their teams comfortable with the idea before AI came along.
The good news is that all of this work makes your business better regardless of what happens with AI. Clean data, connected systems, and efficient processes are valuable whether or not you ever connect an AI tool to them. You're not gambling on AI paying off. You're just running a tighter ship, and if AI does deliver on its promise (and we think it will), you'll be ready.
If you'd like a hand working out where to start, or you want to talk through what AI readiness looks like for your specific setup, give us a shout. No hard sell, just an honest conversation about where you are and what makes sense. That's what we do.
Frequently asked questions
What does AI-ready actually mean for a business?
AI readiness isn't about having the latest AI tools. It means your business data is clean, consistent, and accessible, your systems can connect to each other, and your team understands where AI can save time. Most AI projects fail because the business wasn't prepared, not because the technology wasn't good enough.
What is the first step to preparing my business for AI?
Start by identifying where your team wastes the most time on repetitive, manual tasks. Ask your team leads what tedious work takes up their week, then rank those tasks by how much time they actually cost. This gives you a clear business case for AI and stops you chasing features that don't solve real problems.
Why is data quality so important for AI?
AI tools learn from and work with your existing data. If your data is inconsistent, incomplete, or scattered across disconnected systems, the AI's output will reflect that. Cleaning up your data by standardising entries, filling gaps, removing duplicates, and agreeing on definitions across the business is the single most important step in making AI useful.
Do I need a big IT budget to get AI-ready?
No. The most important preparation work costs nothing beyond time. Auditing your systems, cleaning your data, mapping how information flows between tools, and identifying your biggest time-wasters can all be done without buying anything or hiring anyone. This groundwork also makes your business run better regardless of what happens with AI.
Should I roll out AI across my whole business at once?
No. Pick one high-impact area where you also have reasonably clean data, and run a small pilot project first. Measure the time saved, get honest feedback from your team, and expand from there. Trying to implement AI across the whole organisation at once is a common mistake that leads to expensive disappointment.
