The Conversations That Matter Before the Code
We recently engaged in a project helping one of our clients use AI to optimise how they deployed internal resources. We all discussed and expected the challenge to be data and tooling.
How we were wrong! The real challenge was human.
Everyone was already talking models and automation, no one asked the simplest question: what does “optimised” even mean for us?
That’s when we realised the project wasn’t about technology.
Act 1: The Real Problem Wasn’t the Algorithm
The client wanted to balance workloads across teams and sites. Their systems tracked everything as you would expect, from projects, time, output and it still resulted in people overworked in some areas and idle in others.
We were asked to “use AI to fix it.”
When we looked closer, the problem wasn’t a lack of insight. It was a lack of agreement.
No one had defined what “good utilisation” meant. Who owned the decision to move resources? what was happening was one manager optimising for cost, another optimised for speed and one team 's focus was morale.
We pressed pause and started with alignment.
Act 2: Starting with the Right Conversations
In those first few sessions, we didn’t touch data. We talked.
We asked three questions:
Within a week, we had the first breakthrough. The team realised the issue wasn’t visibility, it was ownership. No one was accountable for resource decisions across departments.
That became the starting point for the AI work.
Act 3: Translating Intent into Computation
Once there was clarity, the technical work became simple. We translated their goal into three data-driven questions:
Recommended by LinkedIn
From those, we built a model to surface mismatches between effort and impact.
The model was just the evidence. The conversation had already solved the problem.
Act 4: Seeing the Human Map Before the Data Map
Another insight came when we mapped how work actually flowed. Emails, Teams chats and meeting patterns told a different story to the org chart.
The people holding the system together weren’t always the ones with titles, they were the connectors.
By visualising that informal network, the company finally understood where to intervene.
Sometimes the fix wasn’t reallocation. It was recognition.
The data reflected what the humans already knew but had never seen in one place.
Act 5: The Quiet Lesson
By the end, the client gained more than an optimisation model. They built a new way of working.
They now start every project with alignment sessions before analysis.
They measure “clarity achieved” before “data prepared.”
And they run story-based prototypes before technical proofs.
Because once the right conversations happen, the maths becomes easy.
We’re still early in this work with our client, but a clear pattern is emerging.
The organisations gaining the most from AI aren’t the ones building the biggest models.
They’re the ones learning to talk differently before they code.
My take away, this is where transformation needs to start, in the conversation, not the computation.
Humans Leading The Loop™ (HLTL)!
