Source: Be Datable, building on Foundation Capital's "Context Graphs" thesis
Memory & Context
The Art of Forgetting: Why organizational intelligence requires curation, not accumulation
Why This Matters for Your Organization
Everyone is racing to build context systems. Capture every decision. Store every rationale. Feed it all to the AI. But that's backwards. Accumulation without curation creates institutional baggage, not intelligence.
The diagram above shows the real challenge: raw context and noise must pass through a filter before becoming actionable signal. Most organizations skip this step. They store everything, hoping AI will sort it out. It won't. AI amplifies both signal and noise.
Forgetting is a feature, not a bug. The question isn't "how much can we remember?" It's "what should we choose to forget?"
The Five Memory Capabilities
1. Persistent Memory
Strategic precedent that should guide behavior indefinitely.
When your board approves a privacy-first policy or your legal team establishes compliance boundaries, those become persistent. They should influence every future decision.
2. The Forget Filter
Separates tactical from strategic, explicit from implicit.
Not every decision deserves preservation. The discount you gave because of a one-time service failure? Tactical. The restructured deal terms that became your new standard? Strategic.
3. Ephemeral Memory
Time-decay algorithms based on access patterns.
A decision referenced weekly stays fresh. One never accessed gradually fades. This mimics human memory. Relevant information remains accessible while irrelevant details naturally decay.
4. Temporal Validity
Tracks when facts are true (and when they stop being true).
A customer's contract value is only accurate as of a specific date. A competitor's pricing is only valid until they change it. Change deltas become signals in themselves.
5. Fact Resolution
Reconciles conflicting signals from multiple sources.
Your CRM says the customer is happy. Support tickets say they're frustrated. Renewal conversation says they're considering alternatives. Which is true? All of them, at different moments. Resolution determines which takes precedence for which decision.
How to Use This Framework
For Your AI Strategy
- Before feeding context to AI, pass it through the forget filter. Not all history is useful history.
- Timestamp everything. AI can't reason about time without temporal markers.
- Build explicit resolution rules. When sources conflict, which wins?
- Include external context. A competitor opening down the street changes everything, but most internal systems miss it entirely.
For Your Knowledge Management
- Distinguish Knowledge Graphs (what happened) from Context Graphs (why it mattered). Both are necessary.
- Long-unchanged information should trigger review, not just exist forever. "You haven't looked at this in 3 years. Is this still how you want to operate?"
- External data brings in what's happening outside your walls. Competitive intelligence, market signals, regulatory changes. Most context discussions miss this entirely.
For Your Team
- Document the "why" alongside the "what." Future you (and future AI) will need it.
- Create sunset rules. When does a decision's rationale expire? Define it upfront.
- Establish source hierarchy. When Sales and Finance disagree, which number goes in the board deck?
Self-Assessment
Answer these questions to evaluate your position.
What decisions from 5 years ago are still influencing behavior today?
Some should be (core values). Many shouldn't be (tactical choices made in different circumstances).
When was the last time you reviewed your "always do it this way" rules?
Stability isn't always good. Sometimes it means everyone forgot to update.
What external signals should change your decisions but don't?
Competitor moves, market shifts, regulatory changes. Are these automatically incorporated?
When different systems give different answers, how do you resolve it?
If you don't have explicit rules, you have implicit chaos.
“Forgetting is a feature, not a bug.”
Without selective memory, organizations drown in their own decision history, unable to distinguish signal from noise. Context dominance requires curation, not accumulation.
Sources & Further Reading
Get more frameworks like this
Weekly frameworks for becoming more data-able in the age of AI.
