Human Possibilities in the Age of AI: Why I Wrote a Book About Where AI Gets It Wrong

I have spent the last several years building merchi.ai, an AI platform that generates product content for retailers. It works. Grosvenor Flooring used it to clear a backlog of 1,000 products and grow their online revenue by 976%. That outcome is real, and I am proud of it.

But I wrote a book because I kept noticing something uncomfortable about how most AI deployments work (including, if I am being honest, how many early conversations about merchi.ai went).

The first thing most people ask when they hear about AI that generates product descriptions is: “So you are replacing the copywriters?” The answer is more complicated than yes or no, and the space between those two answers is where the book lives.

The problem most AI conversations skip

Most businesses are running at a fraction of their human potential. Not because their people are inadequate. Because the system is consuming them.

The typical merchandising team at a mid-sized retailer is not spending most of their time doing the work that actually matters: understanding product ranges, making buying decisions, building supplier relationships. They are spending it on data entry: copying specifications from supplier PDFs into PIM fields, writing descriptions for products that have no copy, uploading images, chasing attributes that arrived incomplete.

This is what I call compensatory work. It exists not because the business designed it to, but because the system has gaps, and humans quietly absorbed them over time. It shows up everywhere: in the extra spreadsheet that bridges two systems that do not talk to each other, in the informal WhatsApp group that coordinates decisions the CRM cannot capture, in the product descriptions written manually because no one automated the process.

AI can absorb this work. That is the genuine promise. Not “do humans’ jobs”, but “stop humans from compensating for a broken system”.

The distinction that matters

Here is the uncomfortable part. The same AI deployment that creates human possibility in one organisation simply reduces headcount in another. The difference is not the technology. It is whether the organisation has the vision and governance to convert freed capacity into something better.

When a retailer deploys merchi.ai and clears a 1,000-product content backlog, two very different things can happen next. In one version, the merchandising team gets its time back and redirects it toward the things that actually drive revenue: better range curation, deeper supplier collaboration, faster time-to-live for new products. In the other version, the headcount goes.

Both outcomes are real. Neither is automatic. The difference is intentionality.

This is what the book is about. Human Possibilities in the Age of AI is a framework for deploying AI in service of people, customers, and businesses, not instead of them.

The Grosvenor Flooring chapter

The Grosvenor Flooring deployment is a chapter in the book for a specific reason. It is the cleanest example I have of what good looks like.

Before merchi.ai, the team was managing a growing catalogue with a content backlog that was limiting their online performance. Products without descriptions cannot rank in search. Products without structured attributes cannot appear in filtered navigation. The backlog was not a failure of effort. It was a structural problem with no manual solution that scaled.

merchi.ai cleared the backlog. But the more important thing is what happened to the team. They did not lose their jobs. They got their attention back. The time that had been going into compensatory data entry was redirected toward the work that actually required human judgment: understanding their product range, building supplier relationships, making merchandising decisions.

976% online revenue growth is the commercial result. The human result is harder to measure but arguably more important: a small team doing work that matters, rather than work that a system should have been doing all along.

What the book argues

The thesis is this: AI implementation should be iterative and pragmatic, but designed and sequenced against a systemic vision, grounded in the jobs customers are hiring you to do, and in how your business actually creates value end-to-end.

This rests on a few propositions that most AI vendor conversations skip past:

Businesses are opaque systems. The real operational logic of most organisations is not in the process documentation. It lives in people’s heads, expressed through informal coordination and daily problem-solving. Before you automate anything, you need to understand what you are actually automating: whether it is compensatory work that should be removed, or valuable informal coordination that would be destroyed if disrupted without replacing it.

The efficiency trap is real. Most AI implementations today make already-fast steps faster, while leaving actual constraints (decision latency, cross-functional handoffs, trust gaps) untouched. They are point solutions applied without systemic vision, creating local optima that collectively underperform.

AI that frees human capacity only creates value if the organisation decides what to do with it. Freed capacity does not automatically become better customer relationships or stronger supplier negotiations. It becomes those things when the organisation is intentional about it.

The new human work is not less work. It is more demanding work. Judgment, relationship, customer understanding, creative problem-solving. These are the things AI cannot replicate and the things most businesses have been systematically under-investing in because their people were too busy compensating for broken systems.

Why this matters for merchi.ai

Building merchi.ai gave me a front-row seat to exactly this dynamic, at scale, across retail.

The retailers who get the most from merchi.ai are the ones who treat it as a system question, not a content question. They configure it to match their exact product data model. They build brand voice into the schema. They think about what the merchandising team does with the time they get back. They treat accurate, structured product data as the foundation for range analysis and purchasing decisions, not just as copy for product pages.

The retailers who get the least from it are the ones who treat it as a content generator. They run a batch, publish the output, and move on. Without the configuration that makes the output accurate, without the review step that catches what the AI cannot know, and without any thought about what the team should be doing now that the mechanical work is automated.

The book is the framework for the first approach. It is aimed at business leaders thinking about where AI fits in their organisation, and at the question of what kind of AI deployment they actually want.

Human Possibilities in the Age of AI is available on Amazon and through the usual booksellers. If you are a retailer thinking about product content and what good AI deployment looks like in practice, the merchi.ai free trial is the practical version of the same ideas.