I have to admit that, like many people, I've been thinking about how to implement AI within organizations while keeping humans in the loop—ensuring they are watching everything AI does, validating, checking, and approving. The human as guardian. The AI as the system that needs minding.
But in my recent conversation with Kieron White of Leading AI, we discussed turning this paradigm around and thinking about AI in the loop.
It's a small shift in language that changes how we think about implementation. Because the reality for most organizations is that their ecosystem of data, information flow, workflows, institutional knowledge, and decision-making processes already exists. And they have people who understand their business deeply. AI is a new capability that businesses and organizations can weave into their existing fabric — not the other way around.
Your Data Is Your Advantage (If You Can Actually Use It)
Most organizations are data-rich and insight-poor. Decades of accumulated knowledge live in documents, databases, emails, and perhaps most valuably, in the heads of experienced people who eventually leave, and so we have what is typically known as 'loss of the institutional knowledge'. According to Gartner, 80% of organizational data exists in unstructured form (i.e., words in documents), and GenAI with Retrieval Augmented Generation (RAG) suddenly makes unstructured data searchable in ways similar to querying databases.
And alongside the mountain of data sits a lot of accumulated inefficiencies. Clumsy processes no one has had time to rethink. Files and documents passed around out of habit. Correspondence chains or processes that exist because "we've always done it this way." Data collected religiously that no one quite knows what to do with — "one day we may look into that."
Trust me, I should know - my organization has thousands of proposals, strategies, and things we've written over the years and forgotten about. Just like us, many organizations produce an incredible amount of information and content all the time. It's impossible — and at times not essential — to act on all of it.
The problem for many organizations is not a lack of data, especially when you consider relevant data outside your organization. It's a lack of capacity to process, connect, make sense of the data and act on it at speed. According to an Atlassian 2025 survey, employees waste 9 hours per week searching for and gathering information.
This is exactly where AI loops in. Not as a replacement for the people and processes that work, but as a way to unlock the value trapped in what you already have. AI is a tool, not the protagonist.
Why the Framing Matters
There is enormous pressure right now to "implement AI.” For many organizations this looks like 'adopting CoPilot' and hoping staff will find ingenious ways to use it. Use cases rarely go beyond summarizing documents/emails and writing support - useful, but not transformational. Some organizations respond by drafting ambitious AI strategies that get discussed in leadership meetings, circulated in polished slide decks — and then filed away as everyone returns to work as usual. The strategy felt disconnected from the day-to-day reality, so nothing changed.
The "AI in the Loop" framing challenges this. Instead of starting with the technology and asking "where can we use AI?"", start with your existing knowledge flows and processes, and ask "where would a new capability make us meaningfully better?"
That's a question your people can actually answer. It grounds AI implementation in the work that's already happening rather than requiring you to reinvent the structures you've built over the years just to fit in a new technology.
When you start from this perspective, practical patterns emerge quickly:
Accelerating existing workflows. Your teams already know what they do. AI can help them do it faster and with fewer bottlenecks — not by changing the work, but by removing the friction within it.
Surfacing hidden connections. Valuable patterns exist across siloed data that no human has the bandwidth to spot. AI can traverse those silos and surface insights that were always there but never visible.
Augmenting judgment, not replacing it. The best use of AI in most organizations isn't to make decisions — it's to give decision-makers better, faster, and sometimes challenging inputs so they can make sharper calls. The expertise stays human. The preparation gets supercharged.
Before evaluating AI tools, map your knowledge flows. Where does expertise get bottlenecked? Where is data going unused? Where are smart people spending their time on tedious, repetitive tasks they're overqualified for?
Better yet — ask them. Ask your people what tasks they dread. Ask them what makes them hate Mondays. Ask what they'd fix if they had a magic wand and an extra day each week. Those answers are your insertion points. That's where AI in the loop starts delivering value immediately, because it's solving problems your people already feel. In municipal government, for example, a 2025 UK Public Sector Efficiency Survey found that 94% of public sector workers face process inefficiencies in delivering citizen services, costing an average of five hours per week per employee in extra work or delays. Meanwhile, McKinsey research shows that public sector productivity has declined roughly 20% since 1995, even as private sector productivity has surged. Permit processing, policy compliance, council reporting, zoning administration — these are areas where skilled public servants spend much of their time on paperwork instead of the community-facing work that drew them to public service. If we can supercharge administration and free municipal staff to focus on planning, policy, and residents, municipalities can retain their best people and deliver better services.
The organizations that will get the most from AI aren't the ones building the flashiest AI systems. They're the ones that deeply understand their own data, processes, and people — and treat AI as a powerful new instrument that listens carefully to every signal inside and outside of your organization.
To learn more about how we help our clients implement AI in their processes, contact: slavi.stevanovic@qatalyst-research.ca
