Twitter Linkedin
Twitter Linkedin
Twitter Linkedin
CASE STUDY

standup Automation

No more “who knows how we fixed this last time?”

How we deployed an agentic knowledge layer that captures what senior consultants know — as they say it, with no change to how the team works.

The problem

A ServiceNow and Salesforce consultancy had a knowledge problem hiding as a staffing problem. Senior architects were losing close to a third of their week to the same repeated questions: the right pattern for an approval flow, the gotcha that burned a client last quarter, the config decision that took three days to work out and five minutes to forget. None of that knowledge was written down anywhere a junior consultant could find it. It lived in Slack threads that scrolled away, in calls nobody recorded, in the heads of people who’d eventually move to the next account — or leave the firm entirely. Every offboarding was a small, permanent knowledge loss. Every new hire started the ramp from zero. The firm didn’t need another wiki nobody would update. They needed the knowledge to capture itself.

The Build

We didn’t ask consultants to change how they work. We built the capture layer around the work they were already doing.

Sit inside Teams.

No new app, no separate tool to remember to open. TacitOS runs where the team already talks.

Learn from conversation.

No new app, no separate tool to remember to open. TacitOS runs where the team already talks.

Answer, with a source.

When a question comes up that the firm has already answered, TacitOS surfaces it and shows exactly where the answer came from, so the consultant can verify before acting on it.

Escalate when it isn't sure.

Below a confidence threshold, it says so and routes to the right person instead of guessing. For a firm whose clients depend on the accuracy of that knowledge, a wrong answer is worse than no answer — so it never fabricates one.

Underneath, this meant single-tenant deployment, governed access, and a hard guarantee that nothing trains a shared model. For a firm working inside client environments, that governance wasn’t a nice-to-have — it was the precondition for anyone senior being willing to use it.

The result

The pilot ran 4–6 weeks against one measure: how often junior consultants escalated to the senior in week six versus week one.

  • One senior consultant's time — freed from being the firm's de facto search engine
  • Zero workflow change — knowledge captured passively, from conversations already happening
  • A permanent recordx — client quirks and hard-won reasoning that now stay in the firm regardless of who stays or leaves

The number either moved or it didn’t. It moved — fewer repeated questions reaching the senior consultant by the back half of the pilot, with the platform already running and ready to extend to a second practice as a configuration change, not a rebuild.

Why it holds up

This is the same principle behind everything we build in the automation and knowledge categories: the highest-leverage AI work rarely looks like AI. It looks like removing a role — “senior consultant as search engine” — that only existed because nobody had built a better way to preserve what one person knew.

This is one example of what we build under AI & ML Services — production automation, sandboxed agents, and multi-model routing, engineered to survive past month two.

Book a discovery call
Shopping Basket