Teach it once. Everyone gets the skill.
PromptQL learns by being corrected. It starts from the context already scattered across Slack, docs, tickets, CRM, and warehouse tables, and on every task it shows its work — the sources it pulled, the assumptions it made.
Correct it once, or pull in the teammate who knows. That becomes shared context: a reusable skill (“exclude test accounts from revenue”), team knowledge, or a semantic-model change.
M
MayaCustomer Success10:02 AM
Acme renews in 3 weeks. Are they a churn risk?
ql
Team AI10:02 AM
Analysing usage data
data
skills
Usage is down 10% from last month.
Analysing support ticket sentiment
tickets
crm
Recent tickets read negative — frustrated, all about exports.
Yes — a risk. @Sam, you've been on the latest tickets — any context for @Maya?
S
SamSupport10:03 AM
It's the month-end Friday export lag. Happens every cycle — they panic, we re-escalate. Not a product problem.
ql
Team AI10:03 AM
Team AI wants to learn
Review and edit if needed, then click "Add to wiki"
Acme's usage dips ~10% every month-end because of a known Friday export lag — it is not a churn signal.
Acme's negative support tickets about slow exports are recurring and operational, not product dissatisfaction.
When scoring churn risk, discount known recurring export-lag patterns before flagging an account.
@ Not sure? Tag someone…
The fastest way to maintain context is to
stop asking people to maintain context.
Nobody wants to update the wiki. Everyone wants their work done.
PromptQL turns that to its advantage: the person who hits missing context has the reason to fix it, right in the flow. Their correction becomes cited, scoped knowledge the next session can use.
Context compounds from real work instead of decaying in a wiki nobody opens. Every task is a chance to teach the system.
Wiki Contributions Per Day
Real wiki-contribution stats from our own use of PromptQL, across a team of 70.
Bootstraps shared context in 60 seconds
- Knowledge
- Skills
- Semantic layer
Maya10:00 AM
Seed the wiki for AcmeCorp
PromptQL10:01 AM
On it — reading Slack, Google Docs, Snowflake, PostHog, Salesforce CRM.
Reading from…
Google Docs
Call transcripts
Snowflake
Consumption data
7 wiki pages created
Suggests context updates as people work
- Capture new context
- Link context
- Prevent context rot
- Capture ambiguity and conflicts
Dana10:02 AM
Pull Q1 revenue by region for the board deck.
PromptQL10:03 AM
On it — assuming revenue = analytics.revenue
Q1 revenue by region — 4 rows
Dana10:04 AM
That table's stale — revenue moved to netsuite.revenue last quarter.
PromptQL10:06 AM
Re-pulled from netsuite.revenue — updated.
PromptQL wants to learn
Review and edit if needed, then click "Add to wiki"
Revenue · source is netsuite.revenue, not analytics.revenue (stale since Q4 2025)
Wikipedia-like operating model
- Easy for non-technical and technical users
- Citations to real work
- Revision history, audit trails & editorial controls
- Notifications on changes, page creations and deletions
Govern with scopes
- Easily handle external, confidential & personal use-cases
- Scopes hold end to end: retrieval, creation & update
- Proven with external users (eg: customers), finance & HR teams
- Granular view & edit control
- Bulk operations to make rapid changes
Access control
One shared wiki, with different connected neighborhoods visible to each scope.
The vibe shift
We made AI a team sport.
Fewer meetings. Fewer status updates. Faster decisions. The work & discussion about the work happen in the same place now — shared AI threads.