Project & Account Intelligence
Building an Operational Intelligence Workspace
Project and account teams often have plenty of documentation. The harder problem is turning it into useful context quickly enough to support the work. This prototype explored whether a dedicated AI-Augmented Work Environment could help project managers work across account knowledge, operational guidance, terminology and workflow instructions without repeatedly reconstructing the picture themselves.
What We Found
The knowledge was there. Using it was the problem.
- —The knowledge existed, but was difficult to use in the moment.
- —Context had to be rebuilt from scratch for every task.
- —Operational language created another layer of friction.
- —A document chatbot would only solve part of the problem.
What the Environment Needed to Handle
Not just search. Context, language and workflow.
The environment needed to handle fragmented documentation, role-specific context, operational terminology that was not documented anywhere centrally, and the reality that project managers rarely start a task from a clean slate.
A chatbot answering questions about documents would only address the surface layer. The deeper problem was that useful context came from combining knowledge from multiple sources — and knowing which sources mattered for which task.
Working architecture — how sources, context and tools fit together.
What We Designed
Three layers, built around the work.
Sources
Project documentation, operational guidance, terminology, workflow instructions.
Workspace
A role-aware environment that assembles context around the task at hand.
Tools
Retrieval, comparison, summarisation and drafting — within boundaries.
How It Worked
The environment assembled context. The person decided what to do with it.
Instead of asking a project manager to search, read and reconstruct before starting the real work, the environment pulled together the relevant knowledge, context and constraints around the task.
The person could then review, adjust and act — with the context already assembled and the sources visible. The AI did the fetching, comparing and summarising. The judgement stayed with the person.
What Changed
Less reconstruction. More time on the actual work.
The time spent rebuilding context before starting a task dropped noticeably. The environment made it easier to find the right information, understand the surrounding context, and move forward without repeatedly asking around.
Equally important: the sources were visible. When the environment assembled context, the person could see where it came from and judge whether to trust it.
What We Learned
The domain was specific. The environment was bespoke. The pattern was reusable.
The specific sources, terminology and workflows were unique to this project. But the underlying pattern — assemble context around the task, keep sources visible, let the person decide — transferred to other domains.
The environment was bespoke. The method was repeatable.
Emerging Bigger Picture
This was one instance of a larger pattern.
Wherever knowledge exists but is hard to use in the moment, the same shape appears: scattered sources, repeated reconstruction, and people carrying too much in their heads. The environment changes what the work feels like — without changing what the work is.
“The domain was specific. The environment was bespoke. The pattern was reusable.”