Yes, I think it addresses a real pain—but the key isn't "connecting tools." Hundreds of products already do that. The real problem is turning fragmented activity into shared organizational understanding.
Most knowledge workers don't struggle because information is unavailable. They struggle because it's scattered across:
- conversations (Slack, Teams, email)
- decisions (meetings)
- plans (Notion, Jira, Asana)
- implementation (GitHub)
- people's heads
When someone asks, "What's blocking the Q3 launch?", they're not asking for documents. They're asking for a synthesized, trustworthy explanation.
The opportunity is to become an organization's "reasoning layer," not just its search layer.
The pain is strongest in a few situations
Cross-functional work
Marketing thinks engineering is waiting on legal. Engineering thinks marketing hasn't finalized messaging. Legal doesn't know they're blocking anything.
New employees
Instead of asking ten people the same questions, they ask the AI:
- Why did we choose this architecture?
- Who owns pricing?
- Why did we delay this feature twice?
Managers
A huge percentage of management is stitching together status updates from meetings, Slack, docs, and dashboards.
Executives
Executives often want:
"Tell me what's changed since Monday."
rather than
"Search my documents."
Where current tools fall short
Today's copilots mostly retrieve information.
The harder problem is reconstructing narratives.
Imagine asking:
"Why are we two weeks behind?"
A useful answer might be:
The schedule slipped after API changes on May 8. Engineering completed implementation, but QA found authentication regressions. Security requested another review, delaying release. Marketing moved launch assets by one week to match engineering's revised timeline. No one explicitly updated the roadmap, which is why different teams still reference different dates.
That's much closer to what humans actually need.
The biggest risk: trust
If the AI confidently invents a reason for a delay, people stop using it.
I'd optimize for showing evidence.
Instead of:
Engineering is blocked by design.
Show:
Evidence
• Alice: "Waiting for final mocks."
• GitHub PR #492 still open.
• Jira ticket marked blocked yesterday.
Confidence: High
People trust systems that explain themselves.
How to reduce meetings without hurting relationships
This is the more interesting challenge.
Meetings do three jobs:
- transfer information
- make decisions
- build relationships
AI can replace (1).
It can partially help with (2).
It should almost never replace (3).
Many organizations accidentally bundle all three together.
Before meetings
Instead of everyone giving updates:
"Here's what I worked on..."
The AI generates:
- progress
- blockers
- dependencies
- unresolved questions
Everyone reads it beforehand.
The meeting starts with:
"Let's discuss the two disagreements."
Instead of spending 30 minutes on status.
During meetings
The AI should quietly help.
Examples:
This topic was already decided three weeks ago.
or
Marketing is waiting on Engineering, but Engineering believes they're waiting on Marketing.
Or:
We've spent 18 minutes discussing a question that already has documented acceptance criteria.
It shouldn't dominate the conversation.
After meetings
Instead of summaries:
Generate:
- decisions
- owners
- deadlines
- assumptions
- risks
- changes since last meeting
Then automatically update:
without someone copying notes around.
Proactive insights are even more valuable
Don't wait for questions.
Surface things like:
Three teams are planning overlapping customer research.
Five engineers independently proposed the same solution.
Sales has requested this feature 14 times this quarter.
The roadmap assumes a dependency that hasn't been staffed.
This feels less like search and more like a chief of staff.
One feature I'd build
Instead of another chatbot:
Organizational Memory Timeline
Every important decision becomes a living object.
For each decision:
- why it happened
- alternatives considered
- people involved
- downstream consequences
- later reversals
Then someone can ask:
Why did we move away from PostgreSQL?
or
Why isn't Europe included in the Q3 launch?
The AI answers with the history, not just the latest document.
Organizations repeatedly lose this context when people leave.
How I'd avoid weakening human relationships
A useful principle is:
Replace coordination, not connection.
Good candidates for AI:
- status updates
- document lookup
- recurring reporting
- meeting notes
- dependency tracking
- action-item follow-up
Poor candidates:
- performance feedback
- conflict resolution
- brainstorming early-stage ideas
- mentoring
- difficult strategic decisions
- celebrations and team bonding
Those conversations benefit from nuance, trust, and shared experience in ways that are hard to automate.
Another design choice is to make the AI escalate rather than absorb interaction. If it detects unresolved disagreement, recurring misunderstandings, or emotionally charged exchanges, it should recommend a conversation between the right people instead of trying to settle the issue itself. Likewise, if it notices a decision has broad cross-functional impact, it can suggest a focused 20-minute decision meeting with the specific stakeholders, rather than allowing uncertainty to linger in chat.
The most successful product in this space may not be the one that answers the most questions. It may be the one that helps organizations spend less time synchronizing information and more time making decisions and building trust. That's a stronger value proposition than "AI that searches across Slack and Notion," because it targets the underlying organizational cost: the constant effort required to recreate shared context.
