Artificial intelligence
AE OS
A weekly read on every project and deal, without the status meeting
AE Studio’s own knowledge about its projects, deals, and people was spread across call transcripts, Slack, documents, code, tickets, and billing data, and basic questions took an hour of digging across tools. AE designed AE OS, the AI knowledge management system it now runs on, which connects those sources into one permissioned layer with a source behind every answer, and each week gives leadership a read on every project and deal with recommended actions.
TL;DR
- Pipelines distill call transcripts, Slack, documents, code and ticket history, time tracking, calendars, and CRM records into a permissioned knowledge graph with a source on every claim.
- Each week AE OS publishes a portfolio view of every active project and deal, with budget, billing, projections, and recommended actions, so leadership gets the picture without a status meeting to assemble it.
- Anyone on the team can ask about a project, deal, or person from the AI tools they already use, within project permissions, and skills on the graph draft handoff documents and kickoff decks.
The challenge
The problem.
AE Studio is a product and engineering studio of about 150 people running dozens of client projects at once, alongside an alignment research group and a sales pipeline. The knowledge that made any of it work lived where it lives in every company: in call transcripts, Slack threads, documents, code, tickets, budget sheets, and the heads of the people who had been on a project longest. Asking whether a project was on track meant opening the sprint board, the budget sheet, and the Slack channel, then messaging the person who was in the standup, and the better part of an hour later having an answer nobody wrote down. Most such questions went unasked, and the answers that were never written down left with the people who knew them.
AE designed AE OS so that this knowledge became something the company could ask and act on. Leadership now gets a weekly read on every active project and deal, with budget, risks, and recommended actions traced back to the call or thread they came from, without a status meeting to assemble it. Any team member can ask about a project or deal they have never touched from the AI tools they already use and get an answer with its sources. Sales walks into prospect calls with the precedents that fit.
It had to work inside AE’s own permission rules. Access is scoped per project, so restricted client data stays out of reach, the knowledge layer can be switched off for a project while its automations keep running, and every fact carries the source it came from.
The solution
What we built.
One Permissioned Knowledge Layer
Pipelines read from the systems where the work already happens, including call transcripts, Slack, documents, code and ticket history, time tracking, calendars, and the CRM, and distill them into a graph of claims about people, projects, decisions, billing rules, and working styles. Every claim keeps its source, so an answer can be traced back to the call or thread it came from, and access follows project permissions.
A Weekly Read on Every Project and Deal
Each week AE OS runs an intelligence cycle over the graph and publishes a portfolio view of every active project, with whether it is on track, per-project pages with budget, billing, and projections, and a sales view of active deals. A program director or salesperson gets the picture without a meeting and can open the evidence when something looks off.
Recommended Actions
For each project and deal, AE OS proposes a short list of actions from a gap and risk analysis, written the way an outside advisor would put them. Model choice is tested against the quality of those recommendations rather than assumed.
Available From the Tools People Already Use
The graph is reachable from the AI tools the team already works in, through a connector, and from an AI assistant in project channels, so nobody has to learn a new tool. On top of it sit skills that do whole jobs, such as drafting a project handoff document from the graph, the SOW, and the call transcripts, or assembling a kickoff deck.
Outcomes
What shipped.
A small internal team stood up AE OS and has grown it one use case at a time, starting with a portfolio view people opened every week and adding deals, budgets, and skills from there.
Leadership reads one weekly view of every active project and deal, with the evidence behind it, without a status meeting to assemble it
Budget, billing, and projection views per project and across the portfolio, replacing manual roll-ups
Weekly recommended actions per project and per deal, which the sales team described as sharp enough to act on
Any team member can ask about a project, deal, or person from the AI tools they already use and get an answer with its sources
Skills on the graph that draft project handoff documents and kickoff decks from what the company already knows
Key takeaways
What we learned.
- The bottleneck was context, not capability. The same model that flounders on a bare prompt gives specific, useful advice when it can read the company’s own history.
- Capture from where the work happens rather than from interviews. Transcripts, Slack, and tickets already hold the procedures and the exceptions, and people asked to write them down leave most of it out.
- A source on every claim is what lets people trust an answer, catch a stale one, and keep restricted data out of the wrong hands.
- Advice and task tracking are different things. Recommendations that read like an outside advisor get read, and a second to-do list gets ignored.
- One use case at a time worked. A portfolio view people opened every week earned the right to add deals, budgets, and skills, and a grand ontology designed up front would not have.
In summary
Bottom line.
AE Studio designed AE OS so that the knowledge scattered across its calls, channels, documents, and systems became something the company could ask and act on. Leadership gets a weekly read on every project and deal with the evidence behind it, every team member gets the company’s history through the tools they already use, and clients get a working example of the AI knowledge management system AE designs for them, because AE hits the edge cases on itself before a client rollout does.
FAQ
Frequently asked.
What feeds AE OS?
How is sensitive information handled?
Do people have to use a new tool?
What does the weekly read contain?
Do you build this for clients?
Let's talk
Bring us the hard problem.
We'll bring the team that ships.