Your Pipeline Is Lying to You: What Business Leaders Really Need to Know About AI

Your Pipeline Is Lying to You (And AI Is Not the Fix)

AutomateNow brought together a small group of senior business leaders at Level 39, Canary Wharf, for an invite-only session focused on one question: how do you actually get value from AI in your business?

Not theory. Not demos. Peer knowledge, shared honestly around a table by people who are testing, failing, learning, and getting results in equal measure.

The format is simple. A guest leader walks the room through how they are using AI in practice. Then the floor opens. Everyone shares what they are seeing, what is working, and what is not. What follows is drawn from that conversation.

The Hype Is Real. The Results Are Not Always.

Russell Dalgleish, who has spent over 25 years working in technology across Europe and the US, opened with a line that landed instantly.

"Artificial intelligence isn't intelligent."

He was not being provocative. He was being precise. The point was not that AI is useless. The point was that we are, once again, in the grip of a hype cycle. And hype cycles, as history shows, are dangerous if you follow the crowd rather than the problem.

"Think about the things that have gone wrong in each of these different heights. Apps. Digital currency. Outsourcing. The pattern is the same every time."

The businesses that benefited from each wave were not the ones who moved fastest. They were the ones who played carefully, kept contracts in place, stayed close to the problem, and avoided handing over control they did not understand.

The message was clear: AI is not a strategy. It is a tool. And tools only work when the people using them know what they are trying to build.

The Real Problem: Your Pipeline Is Built on Bad Data

Pawel Kiezynski, who leads integration and AI implementation work, brought the conversation back to earth with a diagnosis that every sales leader in the room recognised. "Your pipeline is lying to you."

Not because the technology is wrong. Because the data underneath it is wrong.

Most organisations that attempt to implement AI do so on top of a foundation of spreadsheets, manual workarounds, disconnected systems, and inconsistent processes. The AI is technically correct. The output is practically useless. Because the information feeding it is unreliable.

"If your finance system is telling a different story than your CRM, and you build AI on top of one of them, you will get confident answers to the wrong questions."

This is not a new problem. It is the same clarity problem that stalls growth before AI ever enters the picture. AI simply makes it more expensive and harder to ignore.

The path forward, as Pawel described it, follows a straightforward sequence. Inventory your data. Map your processes. Remove the workarounds. Automate the repeatable. Only then does AI earn its place.

"Think of AI as a new team member. You would not hire someone and give them no briefing, no process, no access to accurate information. Why would you do it with AI?"

The Digital Twin Moment

The most memorable story of the morning came from Russell Dalgleish. Sitting in a cafe in Spain, he received a call from a friend who was an HR professional in a difficult spot. He opened his phone, spoke to what he described as his "digital twin" (an AI agent trained on everything he had ever published, every article, every LinkedIn post), and within minutes had a proposal and a brochure ready to send.

The following week, his friend had three inbound requests for meetings.

"I didn't go for process automation. I just played with it."

The lesson was not that AI should run your business. It was that the leaders who benefit most are those who experiment early, stay curious, and keep the human decision at the centre.

Russell's recommendation to the room was blunt: play with it. Pick a problem. Start small. Learn what it does and, just as importantly, what it does not do.

A Warning Worth Repeating

One of the more sobering moments came when the conversation turned to data, privacy, and contractual risk.

A subject access request for a former employee now potentially includes every meeting transcript, every AI-assisted note, every recorded interaction involving that person. The legal implications are still being worked out. The practical risk is immediate.

"Be careful what you wish for when you say play with your digital self."

The point was not to frighten anyone away from AI. It was to reinforce what good governance looks like: contracts with suppliers, clarity about what is being recorded, and an honest assessment of what data you are comfortable sharing.

The leaders who will build lasting advantage from AI are not the ones who automate fastest. They are the ones who build trust, with their teams, their clients, and their data.

What the Room Left With

By the end of the morning, a few things were clear. AI is not going to replace good thinking. It is going to amplify whatever thinking you already have in place, good or bad.

If your processes are clean, your data is reliable, and your team understands the problem they are solving, AI becomes a powerful co-pilot. If those things are not in place, AI becomes a mirror that reflects your confusion back at you with considerable confidence.

The question every leader in that room left asking was not "which AI tool should I buy?"

It was "what do I need to get clear before I go anywhere near it?"

That is the right question. And it is the one worth starting with.

AutomateNow runs invite-only leadership sessions bringing together senior business leaders to share what is actually working in sales, AI, and growth. If you would like to be considered for a future session, visit automatenow.uk.

FAQ

Why do most AI implementations fail?

Because the data they rely on is inconsistent. AI does not fix bad processes. It scales them.

What is the first step before implementing AI in sales?

Audit your data and map your processes. If your CRM and finance system tell different stories, fix that first.

What is a digital twin and should I have one?

A digital twin is an AI agent trained on your own published knowledge, writing, and thinking. It can help you answer questions, draft documents, and explore opportunities faster, using your own perspective as its foundation.

How do I know if my pipeline data is accurate enough for AI?

If your team uses workarounds, exports to spreadsheets, or manually adjusts reports before presenting them, your data is not AI-ready.