AI’s Bigger Prize Is Growth, Not Cost-Cutting

Amin Mrini is Group Chief Commercial AI Officer at Informa, the B2B information services group. He spent close to a decade in strategy, starting in management consulting, before moving into product and digital leadership. Half his job, by his own account, is redesigning workflows around AI.

Most companies are pointing AI at what they already do, in the hope of doing it cheaper. “I think this idea of doing things for cheaper is incredibly short-termist and mono-dimensional as a way of trying to unlock AI value.”

He allows that most will pass through a cost phase first, because saving is the easier thing to measure. In his formulation, a cost saving is banked once, a new product compounds. He wants the productivity equation turned around. Revenue or ARR per head, rather than the number of heads needed to produce each dollar of profit.

“Being protective and thinking about costs only is a defence strategy that will ultimately affect your market share and your growth potential long term. Someone out there is trying to eat your lunch and is thinking new products, new services, new demand, new ways of distributing content, information, products, you name it.”

It matches what aibl sees across mid-market organisations. The firms making progress aren’t the ones with the largest AI budgets. They’re the ones asking what each person can now produce.

The task is the wrong level

“We need to ban this AI use case term.”

Ask for AI use cases and you get a hundred brainstormed ideas and a set of discrete tools, most of them chatbots, built in the hope that productivity rises or 10 to 15% of cost comes out. It doesn’t work.

“A simple example is: I can create presentations more quickly, but if the decision and the meeting during which it takes place is still a monthly meeting, you haven’t saved time. You’ve just given people more time to over-polish an asset or do other things that aren’t unlocking value.”

“The value is only unlocked if you start chaining tasks, which is why we start talking about workflows.”

Start from what humans have to keep

“The question that I ask is what do humans absolutely need to stay in charge of? And then once you protect that or you take that out, you delegate and find ways of giving the rest to AI.”

That inverts where companies usually start, which is what AI can help people do faster or more cheaply. Done his way, it means going through the sequence and marking the points that cannot move. Which steps need a human decision, which need a human approval, which need a person to produce the output before handing it on. Everything left over is available for delegation, subject to accuracy, integration, and compliance.

Making that judgement requires what he calls AI fluency, knowing the process and knowing what the technology can be trusted with. Take invoice reconciliation. The hardest cases are the ones getting pulled out for manual review, so what matters is how far into that pile AI can get. That determines whether productivity actually shifts, and what the team handling what’s left needs to look like. He thinks you find that out by trying things, not by analysis alone.

You can only measure a workflow you decided to change

Token consumption became the proxy for progress, a source of pride that was supposed to show AI fluency. It stood in for a decision nobody had made about what AI was meant to change. “I have an AI strategy. I’ve given everyone GPT or Claude… If you just say everyone’s going to access this free for all, then by definition, you’re not quite sure what you’re trying to achieve.”

What a free-for-all produces is objectives nobody can assess afterwards. Shift everything at once. Be more productive. Do more with less. Stated at workflow level, they become measurable instead. Sales team throughput. Cost of conversion. Pipeline growth of a defined percentage. The same book of business covered by a fifth fewer reps, with the reps it frees up moved onto other work.

Then you need to know where those numbers stood before you started, which he treats as ordinary rather than difficult. Any business launching a new product is accountable for what it does to the customer journey, satisfaction, and financials, and this is the same exercise. Without the baseline there is no uplift to point at.

Attribution will not always be clean, and he is content with a partial view when the effect is large enough to be obvious, which only works if you’ve picked something worth doing. “If you’re starting with the things that are driving the extra 0.01 per cent and that you’re not quite sure within the margin of error, I would argue you’re probably not starting at the right place.”

Put a team behind the workflow, not a tool in everyone’s hands

“Everyone now says reimagine the workflow, reimagine the workflow. It can’t be an individual contributor’s responsibility day in, day out. No one wakes up and thinks, how do I rethink my job?”

“The recipe is just so obvious, it almost pains me to have to say it.” His answer is small technical teams working inside the functions themselves, embedding AI into operations rather than issuing tools and waiting. In his account that sidesteps the adoption problem and the change management problem in one move, and keeps the ability to rewire whole sequences rather than individual tasks.

The mandate comes from the CEO, including the investment and what the business expects AI to unlock. The AI function itself does not own outcomes across the business. Each function stays responsible for what the re-engineering is meant to deliver, which requires some AI fluency in functions that are not technical.

The engineering is the part he thinks gets underestimated. Invest in it, put a small number of engineering and product people inside the function, and drive the change from there. The AI should be invisible, embedded in the work rather than arriving as another four or five tools to open. “The less AI you see, the better for productivity.”

The role most exposed is the one that only coordinates

Redesign enough workflows this way and the shape of the organisation changes. The pressure lands not on the capable generalist but on what he calls the pseudo-generalist, the person coordinating and managing without deep domain expertise of their own.

That role made sense when execution was expensive and expertise sat in silos. Someone had to sit above it, scheduling the meetings, doing the reporting, making sure the handovers happened. Silos are getting closer together and in some cases merging, and execution is no longer scarce.

Two things gain value. The depth to catch what AI has got wrong and push its output further, and the breadth to cover ground that used to be divided between specialists. The profile he thinks will thrive combines both. His example is a product manager technically fluent enough to prototype, look at design and research, and take on launch planning. “Why would you need a UI designer, a product designer, an engineer, a sales enablement specialist and a marketing manager?” Five or six roles compressed into two or three. “That’s how I’ve built my team.”

What’s left exposed is the lubricant between layers. He is careful about how fast that plays out, though. In his view hybrids can rise while middle-management roles persist for some time yet, and he treats it as a thesis about to be tested rather than a settled one.

Finding those people is the hard part. “I wish I had the recipe.” Hiring one makes the second easier, because people with that skill set recognise it and want to work near it. But he doesn’t think a company can start from a clean slate and recruit a full team of them. Find a small number and use them to raise everyone else. “I’m pretty sure that upskilling is more important than hiring.”