A consultant running a high-stakes strategy session is doing two jobs at once. The one the client can see, and the one they can't.
On the surface: listening, questioning, guiding. Underneath: absorbing everything being said, cross-referencing it against prior work, hunting for the right reference document, forming the next question, and somehow writing all of it down at the same time. For most consultants, the note-taking and document-hunting is what actually degrades the conversation. You can't be fully present when part of your attention is on the keyboard.
That was the problem we were handed. Build something that fixes it. What that actually meant took longer to understand.
The Dual-Task Penalty: Every cognitive resource spent on admin during a live session is stolen from the client. Senior consultants knew this acutely, they described the feeling of "losing the room" while trying to find a document they knew existed somewhere.
Reactive AI Doesn't Work Here: Early thinking pointed toward a chat interface. Type a question, get an answer. We killed that idea fast. Typing a prompt mid-meeting means breaking eye contact, losing your thread, and signalling to the client that you're not fully with them. A consultant can't use a chatbot during a pitch.
The Admin Spiral: Post-session, the burden shifted but didn't shrink. Notes needed structuring, follow-ups needed drafting, references needed filing. Hours of administrative work compounded after every meeting.
We designed the layout before we designed any feature
The workspace is split into two distinct hemispheres, and that division was the first and most important design decision. Left side: static, known, predictable, the agenda, the live transcript as it builds word by word, the session structure. Right side: dynamic, AI-generated, surfaced in real time, extracted insights, flagged topics, suggested references, recommended follow-ups.
The logic was simple: the left side is what the consultant already knows is happening. The right side is what the system is noticing that the consultant doesn't have bandwidth to notice themselves. Two very different cognitive jobs, separated into two columns, so neither contaminates the other.
We spent significant time on information hierarchy within the right panel. Everything the AI surfaces is ranked by confidence and relevance to what's being said right now, not what was said five minutes ago. Stale suggestions fade. Relevant ones come forward. The panel has to feel like a smart colleague reading the room, not a search results page.
The transcript interface required more care than we expected
Live transcription sounds like a solved problem. In a consultation context it isn't. Consultants speak over each other, clients interrupt, there are long pauses, there's crosstalk. The transcript view needed to handle all of that without becoming unreadable mid-session.
Speaker demotion, the ability to mark a segment as background noise or sidebar conversation, was one of the iterations that emerged from testing. Without it, the transcript filled with fragments that the AI would then try to interpret. With it, the consultant can keep the signal clean in real time with a single tap, no typing required.
The transcript also had to feel calm. No flashing, no aggressive animations as new words appeared. Text rolls in at reading pace. The goal was for the consultant to be able to glance at it briefly for confirmation, not read it actively.
Silent Guidance was the feature that changed everything in testing
The mechanism that made the biggest difference wasn't transcription or insight generation, it was a small discreet input that let consultants flag something for the AI without speaking or typing. A brief tap on a keyword in the transcript, or a subtle gesture, tells the system: expand on this. Surface related material. Track this thread.
In simulated testing, this changed how consultants interacted with the session entirely. Instead of mentally bookmarking things to chase up later, they could hand the flag to the system in the moment and stay in the conversation. The AI would then build on the flagged thread in the right panel, quietly, while the session continued.
That feature, Silent Guidance, was not in the original brief. It came from watching consultants gesture toward their screen during testing, trying to interact with something that wasn't there yet.
The first build was good at capturing. It wasn't good at supporting.
We had a system that recorded everything, transcribed accurately, and generated a solid post-session summary. Consultants liked it. But the real-time value, the thing that would justify having a screen open during a meeting at all, wasn't there yet.
The Live Recommendation engine came out of that gap. By processing the transcript at near-zero latency rather than in batches, the right panel started surfacing relevant prompts and historical data mid-conversation rather than after the fact. The shift from batch processing to streaming was technically significant but the user experience change was dramatic. The tool went from a recording assistant to a live strategic partner.
What the project confirmed: AI integration in high-pressure environments has to be invisible
Every interaction model we tested that required the consultant to actively query the AI failed in live session conditions. The cognitive overhead of forming a prompt is too high when you're also managing a client relationship in real time.
What worked was a system that made decisions about what to surface without being asked, and was right often enough that consultants started trusting it rather than second-guessing it. Building that trust took iteration after iteration of tuning the relevance model. Getting it to 80% accuracy on surfacing useful references felt like a breakthrough. Getting it to feel calm and confident at 80% accuracy was the harder design problem.