The real barrier to AI transformation may not be AI itself. It may be the tools AI is supposed to transform.
Picture this. An employee opens the CRM — the customer relationship management tool — to update a prospect record. They navigate three screens, fill in seven fields, select options from two dropdown menus, validate, wait for the page to load, move to the next screen. Three minutes for an operation that AI, if it could access the system directly, would handle in a fraction of a second.
Meanwhile, ChatGPT is open in another browser tab. They copy and paste information back and forth. The AI does its job — summarising, analysing, rephrasing. But it does so outside the system, because it cannot get in.
This is not an AI problem. It's a software problem.
A design assumption that became invisible
To understand what's at play, you need to go back to the origins of our tools.
Since the 1980s and 1990s, enterprise software has been built around a concept engineers call the HMI — the human-machine interface. As the name suggests, this interface was designed for a dialogue between two actors: a human and a screen. The human inputs. The software processes. The entire architecture — windows, forms, required fields, menus, step-by-step workflows — is built around this assumption.
Whether we're talking about an ERP (enterprise resource planning system, centralising financial, logistics, and operational data), a CRM, accounting software, or a project management tool, the logic is the same: the software guides a human through a sequential process.
This model made sense when humans were the only ones capable of understanding context, making choices, and taking decisions. It shaped forty years of business software.
But it is becoming a bottleneck.
AI is not an assistant. It's a second user.
Most talk about AI in business presents it as an assistant helping humans use their tools better. A chatbot bolted onto the CRM. An automatic summary in the inbox. A copilot in the spreadsheet.
That's true, but it's not enough. What is changing at a fundamental level is that AI is not merely an assistant to the human. It is a second user of the system — capable of reading and interpreting the software's data, proposing or making decisions, writing to the system, and doing so at a speed, volume, and consistency no human can match.
But this second user doesn't know how to fill in a form. It doesn't click buttons. It doesn't wait for a screen to load. It needs what developers call APIs — application programming interfaces, essentially doors designed for machines to communicate with each other. It needs structured, accessible data. Real-time data flows.
And most tools in place at SMBs offer none of this.
The symptoms you already recognise
This mismatch produces situations every business leader will recognise:
- Double data entry: teams spend hours manually transferring information from one tool to another because systems don't talk to each other
- AI running "alongside" the software: employees use ChatGPT in one tab and their business tool in another, copying and pasting between the two
- Rigid workflows: the software enforces a ten-click process where AI could handle the operation in a fraction of a second
- Trapped data: information exists, but it's locked inside application silos
- Impossible end-to-end automation: you can automate fragments of a process, but never the complete chain, because the central link only speaks to a human
What I discovered by living it myself
I'm not speaking about this subject theoretically. I've lived it.
A few months ago, I made a decision that many would find counterintuitive: I abandoned several software solutions I was using daily — recent tools, high-performing, well-designed — to build my own solutions.
What guided my decision was something very specific: I wanted AI and me to both be able to interact with the system.
With the tools I was using, however modern, I was the only one who could act within the application. AI stayed outside. It could help me prepare what I was about to enter, or analyse what I extracted. But it could not get into the system directly.
By building my own solutions, I was able to design from the outset an open architecture — where AI is not a spectator but a co-user. In practice, AI accesses the same data I do, can query it, supplement it, structure it, trigger actions. And it does so in real time, alongside my own work.
The daily efficiency gain was immediate and substantial.
It wasn't AI that had a problem. It was the tools that weren't ready to accommodate it.
This is not a technical issue. It's a strategic one.
Many leaders would instinctively file this under IT. That would be a mistake.
What's at stake is a company's ability to integrate AI not as a gadget sitting next to its processes, but as a full participant in how it operates.
According to Goldman Sachs, 42% of SMBs say they lack the resources or expertise to deploy AI properly. And among the barriers, tool compatibility is an obstacle that surveys underestimate.
Yet growing SMBs already stand out: they are twice as likely to have a connected technology ecosystem as declining SMBs — 66% versus 32% (Salesforce, December 2024).
The questions to ask are those of a leader: what information system do we want in three years? Are our current tools compatible with the augmented model we're aiming for? Are our tools ready to have two users — or only one?
Anticipate the gap before it catches you
The gap between AI's capabilities and the limitations of existing tools is going to widen very rapidly. Every month, AI advances. Software evolves at its publisher's pace — often too slowly.
Not asking this question today means risking ending up tomorrow with a powerful but unusable AI — like a racing engine in a car whose steering wheel wasn't designed for it.
Understanding this gap, assessing the state of your own systems, and anticipating the changes needed — that's exactly the kind of thinking we support with IMPAICT.
Sources
- Author's personal experience
- Microsoft / LinkedIn, 2024 Work Trend Index (May 2024, 31,000 people, 31 countries)
- Salesforce, Small & Medium Business Trends Report, 6th ed. (Dec. 2024, 3,350 leaders)
- Goldman Sachs, 10,000 Small Businesses Voices (Aug. 2025)
- WEF, Future of Jobs Report 2025
Comments (0)
Be the first to comment.