Last month I sat down with a group of technical leaders in Manchester for a roundtable on AI adoption, hosted by Nimble Approach. It was a wide mix of businesses: firms handling complex legal matters, large consumer-facing organizations, data specialists, logistics companies and consultancies. Our discussion followed two tracks: driving adoption inside a business as part of wider change, and showing customers the value an AI product brings them.
The two tracks had more in common than I expected. Whichever side of the table you sit on, three questions kept coming up.
1. What problem does it actually solve?
Adoption starts with a validated use case. People need to understand what a tool is for, whether it is a general assistant such as Copilot or a product built for one specific job. Without that, you have a solution looking for a problem.
There is a human side to this that I believe gets underestimated. Whether a tool comes from an internal transformation team or an outside vendor, end users worry it is there to replace them. The view around the table was clear: AI lands best when it adds to someone’s role, or takes away the work that gives them little satisfaction. That might be the endless short, repetitive queries, or the long, drawn-out back-and-forth. It holds as true for an employee doing complex work as it does for a help desk handling queries at scale, with or without deep domain knowledge.
Time saved at the macro level is rarely what wins an individual over. Three minutes saved once a day barely registers. What people notice is the kind of work they no longer have to do: the tedious search, the repeated question, the information that is never to hand, the task that drains energy before the real work starts. Take that away and the job feels different. That psychological shift, not the time sheet, is what turns a skeptic into an advocate.
2. Is it worth it to the business?
Individual enthusiasm is not the same as organizational adoption. For a business to commit, value has to be visible to the senior leadership team and measured in output and efficiency.
Finance will ask the obvious question: what does it save for what it costs? I believe the sharper question is opportunity cost. If AI takes on this work, what else could our people be doing that creates more value? Could it open up new markets, products or services, or let us take on more of what we already do well?
At business scale, small gains add up even when no individual feels them. Those three minutes a day come to around 11 hours a year per person, or more than 5,700 hours across a 500-person business. But the business only banks that gain if its people engage, and that brings us straight back to the first question. This is where most AI adoptions fail in my experience. People return to old habits, the tool goes unused, the chatbot goes unanswered, and the business writes off the investment.
3. Is it safe, and is it allowed?
The third question is more binary. Either the technology meets the organization’s security, governance and compliance standards or it does not. Is it ethical and within current legislation? Where does the data live, and who else can see it? What happens to GDPR obligations and the right to be forgotten if the model trains on your data?
An AI product either passes these tests or it fails them. The best ones go further and strengthen an organization’s compliance position rather than merely satisfying it.
There are, of course, many use cases for AI within cyber security itself. In an increasingly complex threat landscape, AI enabled tools could act faster than people, and keep humans in the loop. They can monitor threats and respond to them around the clock, based on rules given to the AI engines, without the need for a 24/7 security operations centre (SOC). However, our focus at the roundtable was on the wider business impact of AI rather than specific tooling.
Where ARK fits in AECO
All three questions matter most in architecture, engineering, construction and operations. Projects are large and complex, the record is spread across drawings, specifications, BIM models, RFIs, emails and contracts, and the data is commercially sensitive. That is the context we built ARK for and the three questions are the ones we hold ourselves to.
- A clear use case that supports people. ARK gives quick answers from your project record, each cited to the document, page and clause. It takes on the searching and cross-checking so people can focus on judgement. FMI and Autodesk research puts time lost to searching, resolving conflicts and rework at 35% of the working week.
- Value the business can measure. Beyond answers, ARK is growing to monitor projects and changes, flag risks and track contractual deadlines. It will then act as an agent within large, complex projects, with skills shaped around the industry. Less searching and less rework means less waste.
- Controls compliance teams can say yes to. ARK is grounded in truth and only works from the data you give it, which reduces, and aims to remove, the risk of hallucination. You choose where your data resides: our cloud, your cloud or on-premise. It stays private, and your documents and queries are never used to train models. A retrieval-augmented generation (RAG) engine designed specifically for AECO underpins it all, complementing those controls. A full audit trail logs every query and answer, recording who asked, when, and the sources behind the response, so you can always show why an answer was given. ARK is also aligned with the accreditations buyers expect, including SOC 2 Type II and ISO 42001, so it can pass corporate governance checks.
- Human in command. ARK supports decisions rather than making them. Every answer is cited so people can check it, and any action ARK proposes waits for a person to approve it. Accountability stays where it belongs, and people can trust the tool because they remain in control.
Adoption is earned
The most impressive demo doesn’t win AI adoption. Answering three questions well does: what does it solve, what is it worth, and can we trust it? If you’re working through those questions on your own projects, I would be glad to continue the conversation. Get in touch.
Peter Willingham
Chief Technology Officer, Ageiro
Peter brings 25 years of commercial software engineering to Ageiro, most recently as Chief Technology & Product Officer at Autocab, an Uber company, where he led product engineering for a SaaS platform operating in 35 countries.