There's a post going round the AI automation forums that's worth more than most of the conference talks this year. Someone who says they've made over £350,000 in AI consulting put it bluntly: the money isn't in agents. It's largely in data plumbing.
Their point was that after months of foundational tooling and infrastructure, only then did the genuinely agentic work begin, and that people think they need agents when what they're actually looking for is a better understanding of what's happening in their business at any given time, and how to connect the dots between those data points.
We build AI for UK businesses for a living, and we can tell you this is the truest thing said about the field all year. It also explains why so many 'AI agent' projects quietly disappear a few months after the impressive demo.
Everyone asks for the robot
The request almost always arrives the same way. A business owner has seen a demo of an AI agent booking appointments or answering customers, and they want that. It's understandable: the agent is the visible, magical part. It's the bit you can show your mates.
But when you sit down and ask what problem they're actually trying to solve, it's never really 'I want an agent'. It's 'I don't know which of my leads went cold', or 'my quotes live in three different places', or 'I can't tell you how much work is in the pipeline right now without asking four people'. None of those are agent problems. They're plumbing problems.
What 'data plumbing' actually means
Plumbing is a good word for it, because like real plumbing, it's invisible when it works and catastrophic when it doesn't, and nobody wants to pay for it until the kitchen floods. In practice it means three unglamorous things:
- 1Getting data out of the places it's trapped: the inbox, the spreadsheet, the WhatsApp group, the CRM nobody updates, the notebook in the van.
- 2Cleaning it up so it actually agrees with itself: the same customer isn't spelled three different ways, a 'lead' means the same thing in every system, dates are dates.
- 3Connecting the systems so information moves on its own, instead of a human copying it from one screen to another.
That's it. It's not clever. It's the digital equivalent of making sure the pipes connect and the water flows the right way. And it is, reliably, where the time, the cost, and the actual business value sit.
Why the money is there
Three reasons, and they compound.
1. It's necessary before anything else works
You cannot automate a process you cannot see. An agent that's supposed to follow up on cold leads needs to know, reliably, which leads are cold. If that information is scattered across an inbox and someone's memory, the agent has nothing to stand on. The plumbing isn't a prerequisite you can skip to get to the fun part. It is the thing that makes the fun part possible.
2. It's ongoing, not one-and-done
A shiny agent is a project. Connected, trustworthy data is a state you have to maintain as the business changes, adds tools, hires people and shifts how it works. That's not a bug in the business model, it's the reason the value is durable. The plumbing keeps earning.
3. It's where trust is built or destroyed
The moment an automation sends a customer the wrong price, or chases someone who already paid, trust evaporates and the whole thing gets switched off. Almost every one of those failures traces back to bad or disconnected data, not to a bad model. Get the plumbing right and the system earns its keep quietly for years. Get it wrong and no amount of clever AI on top will save it.
The part everyone skips: an agent on bad data is worse than no agent
This is the bit the demos never show you. A human looking at a messy spreadsheet knows it's messy and treats it with suspicion. An AI agent looking at the same messy spreadsheet does not. It will confidently act on it. It will email the wrong customer, quote the wrong price, or tell someone their appointment is on a day it isn't, and it will do it faster and more consistently than a human ever could.
Putting an AI agent on top of disconnected, dirty data doesn't save you time. It just lets you make mistakes at scale, with total confidence, out of hours.
This is why the sequencing matters so much. The plumbing isn't the boring thing you do before the valuable thing. It's the thing that determines whether the valuable thing helps you or embarrasses you in front of your customers.
What this looks like for an actual small business
Big consultancies talk about this in terms of data lakes and infrastructure. For the businesses we work with, it's much more down to earth. A few real examples of 'plumbing before agents':
| They asked for | What they actually needed first |
|---|---|
| An AI agent to answer customer enquiries | Every enquiry (calls, web, WhatsApp, Instagram) landing in one place instead of five |
| An AI agent to follow up leads | A single, reliable list of who's a lead and what stage they're at |
| An AI to write our quotes | Their prices and job types written down in one place instead of in the owner's head |
| A bot to chase invoices | Their job-completion data connected to their accounting software |
In every one of those cases, the connecting work delivered most of the value on its own, before a single 'agent' existed. Once enquiries all land in one place, a human can answer them fast. Once the lead list is reliable, follow-up becomes possible. The agent, when it eventually arrives, is the cherry on top of a cake that's already worth eating.
So should you forget about agents?
No. That's the wrong lesson, and the original post was careful to say the agentic work does eventually come, and it's real. The lesson is about sequence, not about avoidance.
Agents are genuinely powerful once they're standing on connected, trustworthy data and given clear boundaries. At that point they stop being a demo and start being an employee who never sleeps. But that's the last 10% of the work, and it only pays off because of the 90% underneath it. Anyone selling you the 10% while quietly skipping the 90% is selling you a very expensive way to make confident mistakes.
How we approach it
This is exactly why we start small and refuse to sell twelve-week transformations. The first thing we build for a business is almost never an agent. It's usually a piece of plumbing: getting every enquiry into one place, or connecting two systems so a human stops retyping data between them.
That first piece earns its money on its own, it proves we can be trusted with the next one, and, quietly, it lays the foundation that a genuinely useful agent will one day stand on. By the time we build the agent, it works, because everything underneath it already does.
If you want to know which piece of plumbing would move the needle most for your business, our free audit reads your website and tells you where the manual, disconnected work is, in about 30 seconds.