AGI Scarce Resources: Practical Opportunities & ReadyCall Application

· business, product-design, personal-growth

Companion to: [[2026-02-06-what-remains-scarce]]

The personal, applied dimension of the AGI scarce resources discussion. What to actually build, who you need to be to build it, and the honest assessment of ReadyCall.

Five Types of Context Worth Collecting Now

Not all data can wait. Some has a collection window — miss it and it's gone forever.

What makes context "must collect now": the information will disappear (conversations fade, people forget, transient states expire), value comes from time-series completeness (gaps can't be backfilled), or the collection opportunity is tied to a specific event or behavior (can't be recreated after the fact).

1. Expert Tacit Knowledge (Most Urgent)

The reasoning process behind expert decisions — what factors they weighed, what they excluded, what trade-offs they made, why they chose A over B. Results get recorded (which contract was signed, what drug was prescribed). The reasoning does not.

Why urgent now: Baby boomer experts are retiring across every industry. An experienced building estimator who can glance at blueprints and estimate costs within 5% — that skill took 30 years and thousands of projects to build. When they retire, the reasoning disappears permanently. It's not in any textbook or database because tacit knowledge is inherently hard to articulate.

Collection method: Not asking experts to write knowledge documents (they can't — tacit knowledge resists explicit expression). Instead, embed tools in their workflow that capture context while they work — what information they looked at, where they paused, what they decided, what the outcome was.

Priority domains: Building estimation, insurance underwriting, factory QC engineering, medical diagnosis, procurement negotiation.

2. Disappearing Industry Ecosystem Knowledge

Entire industry knowledge graphs are dissolving as markets restructure.

Example: China manufacturing supply chain maps. Two decades of accumulated knowledge — who subcontracts to whom, real capabilities vs. claimed ones, reliable partnerships, cost structures — lives in the heads of veteran sourcing agents and trading companies. As supply chains shift to Southeast Asia, these agents retire or switch industries. In five years, many factories will have closed or pivoted, the sourcing agents will have moved on, and this relationship map is permanently lost.

Example: Traditional media buying expertise. As advertising goes programmatic, the generation of media buyers who know each TV station's real viewership, each magazine's actual influence, each outdoor location's true foot traffic — they're leaving the industry. Their experiential knowledge exists in no database.

3. Human Behavioral Preferences & Decision Patterns

How people actually make decisions vs. how they say they do. This data doesn't "disappear" but its value comes from time-series length — the earlier you start, the richer the profile.

Change trajectories are more valuable than snapshots: "this person recently started preferring quality over price" reveals a life-stage shift (income increase? role change?) that a static label misses.

Caveat: Privacy sensitivity is high (requires explicit user consent), and general agent platforms (Apple, Google) have structural advantages in breadth. To compete, must go deeper in a specific vertical — e.g., procurement professionals' decision patterns during supplier evaluation. A general agent won't specifically track "procurement decision style," but a procurement tool can.

4. Physical World Baseline Data

Some physical-world data must be captured now because the physical world itself is changing, and "before" states can't be reconstructed.

Climate baselines: Current flood frequencies, soil conditions, crop yields in specific regions. As climate changes, these baselines shift. Without historical records, you can't quantify "how much has changed." Critical for insurance pricing, agricultural planning, infrastructure investment.

Urban space usage patterns: Post-pandemic hybrid work is reshaping commercial real estate usage, foot traffic, commute patterns. These patterns are still forming. Continuous recording creates long-term value for real estate investment, retail site selection, urban planning.

Business case challenge: Who pays for baseline data? Insurance companies, investment firms, government planning agencies — but these are long-cycle monetization paths.

5. Organizational Process & Decision Knowledge

Every organization has massive undocumented process knowledge — "who actually approves this," "how we resolved this issue last time," "the real decision flow vs. the org chart." This lives in veteran employees' heads, scattered Slack threads, and email chains.

As employees turn over, knowledge continuously leaks. New hires spend enormous time rediscovering what predecessors already solved.

Collection method: Extract and structure from existing work artifacts (Slack conversations, email threads, meeting recordings, ticket resolution histories). Not a new data capture effort — it's mining knowledge from workflows that already produce digital traces.

Six Medium-Sized Opportunities

Brainstormed for someone with product/design background, bilingual English/Chinese, building AI productivity tools, with Philosophy 2 aspirations (product intuition + fast iteration). Medium-sized = not tiny verticals, not coding-agent-scale competition.

1. Post-Meeting Execution Engine

Beyond meeting notes into the action layer. Not "here's what was discussed" but "here's what needs to happen, it's been created in your task system, and I'm tracking whether it gets done." The shift from passive recording to active execution management.

Challenge: "action items" are too diverse across meeting types. Needs a specific vertical to focus (see ReadyCall analysis below).

2. Cross-Border Communication Intelligence

Leveraging bilingual advantage. Not translation (commodity) but cultural-business context bridging. Understanding that when a Chinese factory says "we'll try our best" (我们尽量) it often means "this is going to be difficult/unlikely" — and surfacing that context to the Western buyer.

Specific form: tool that sits in cross-border email/WeChat/WhatsApp communication, flags cultural misalignment, suggests clarifications, tracks commitment differences between what was said and what was meant.

3. Sales Preparation Automation

Every B2B sales rep spends 1-2 hours before each meeting researching the prospect — reading their website, LinkedIn, recent news, checking CRM history, preparing talking points. AI can compress this to 5 minutes.

Existing competition: Gong (post-call analysis), LinkedIn Sales Navigator (research). The gap is in pre-meeting synthesis — combining company research, contact history, industry context, and competitive intel into a personalized briefing.

4. Solopreneur Operating System

Freelancers and solopreneurs juggle clients, invoices, contracts, scheduling, communication across 5-10 separate tools. An AI-native workspace that consolidates: "here are your active projects, this invoice is overdue, this client hasn't responded in a week, you have 3 hours free Thursday for the proposal you promised."

Growing market (freelance economy), underserved by tools designed for teams, product-design-heavy opportunity.

5. AI Tutoring with Real-World Feedback Loops

Current AI tutoring: ask a question, get an explanation. Missing: spaced repetition based on actual understanding, adaptation to learning style, connection to real-world application.

The feedback loop matters: student attempts problem -> AI observes approach and errors -> adjusts teaching strategy -> student tries again -> improved outcomes feed back into method optimization.

6. Complex Document Understanding

Legal contracts, insurance policies, government regulations, technical specifications — documents where "reading" isn't enough, you need "what does this mean for my specific situation?" Current AI can summarize. Missing: personalized interpretation given your context, comparison across documents, extraction of obligations and deadlines.

Startup Philosophies

→ Full analysis in [[2026-02-06-three-startup-philosophies]]. Three archetypes: Structural Advantage (defensible moats, patience), Ride the Wave (speed, product intuition), Enabling Infrastructure (developer ecosystem). Self-assessment: identifies with Philosophy 2 but capabilities still building. Key insight: muscles built by shipping, not by analysis.

ReadyCall: The Honest Assessment

Meeting Notes as a Pain Point

Meeting notes/action items is a real but mild pain point. People forget follow-ups, miss CRM updates, lose action items — but they've survived with workarounds (memory, scribbled notes, colleague reminders). "Good enough" workarounds are a product's worst enemy — users won't actively seek a solution.

The CSM Entry Point (Deep Analysis)

Customer Success Manager emerged as the strongest entry point during analysis. The logic:

Why CSM looks attractive:

  • Predictable action item structure — nearly every client meeting requires: summary email to client, CRM notes update, internal Jira/Linear ticket if product issues surfaced, follow-up reminder
  • Quantifiable failure cost — CSM forgetting a client issue = client asks about it next meeting, CSM is blank, trust erodes, NPS drops, potentially churn
  • Tool gap — Gainsight does CS platform but its meeting execution experience is poor
  • Natural distribution — CSM's follow-up emails can carry product branding; CS communities (Gainsight community, CS Slack groups) enable word-of-mouth
  • Data flywheel — cross-meeting context accumulates ("this client has raised the same feature request 5 times in 3 months"), creating switching costs

The CSM Challenge (The Honest Part)

Clients rarely churn because of CSM meeting execution. Real churn drivers, roughly rank-ordered:

  • Product doesn't solve the problem / insufficient perceived value
  • Product quality or stability issues
  • Customer's own changes (budget cuts, reorg, acquisition, strategy shift)
  • Better competitor
  • Pricing
  • CSM-related factors (slow response, unmet promises, feeling neglected) — maybe 10-20% of churn causation

A client won't churn because CSM forgot one follow-up email. They churn because the product isn't good enough, the price is too high, or their business changed. Good CSM execution improves retention at the margin but is not the determinant.

The attribution problem: If you pitch "reduce churn with better CSM execution," the VP of CS asks "how do I know churn dropped because of your tool vs. the product team fixing that big bug?" Churn is multi-causal. Isolating your contribution is nearly impossible.

Recalibrated value proposition:

  • "Save CSM time" — each meeting saves 15-30 min of note-taking/CRM updating/email drafting. 4-5 meetings/day = 1-2 hours saved. Real but hard to price high — "time saving" tools get low willingness-to-pay.
  • "Look more professional" — every meeting produces a structured summary with clear next steps. Improves client experience. But the effect is soft and unquantifiable.

Both are real. Neither is sharp enough to make a VP say "I need this today."

Rethinking: Where Meeting Execution Failure Actually Hurts

Instead of starting from roles, start from "when does meeting execution failure cause severe consequences?"

Compliance-required meeting records — Medical consultations (EHR requirements), financial advisory meetings (MiFID II/SEC), legal client meetings (billable hours + discussion records). "Didn't record" = "violated regulation, possible fine or lawsuit." Pain goes from mild to acute. But regulatory requirements mean high product compliance bar (HIPAA, SOC2).

Meetings involving financial commitments — Sales negotiations with price/term commitments, procurement negotiations with specific conditions, contract discussions with agreed modifications. "You said you'd give us 10% off" — "We never said that." An auto-record-and-confirm tool has clear value here. But this is "commitment management" more than "meeting notes."

Multi-party project execution meetings — Construction coordination, film production meetings, large event planning. Many independent parties, each assigned specific tasks, failure by one party cascades to entire project. But these usually already have project management tools (Procore, Monday.com).

The Broader Question

Maybe the question isn't "which role's action items hurt most" but whether ReadyCall should center on "meeting execution" at all.

If "meeting-after action items" is broadly mild pain, alternatives:

  • Find the narrow vertical where pain is acute (compliance recording, financial commitments) — but this makes ReadyCall a vertical product with a smaller market
  • Redefine the core value away from action items — "too many meetings" (help decide which aren't worth attending), "low quality decisions in meetings" (real-time data/context during discussion), "repeated discussions" (remind that this topic was resolved 3 weeks ago with X conclusion)
  • Accept meeting notes is a feature, not a product — embed it as part of a larger AI workbench (CSM workbench, sales workbench). But then you're competing with Gainsight, Gong directly.
  • Stop analyzing, start shipping — Philosophy 2's core lesson. Maybe the best entry point isn't found through frameworks but by watching someone work and noticing the moment they curse under their breath.

Granola/ReadyCall's Unique Position

Granola isn't pure Philosophy 1 or 2. It's trying to define a new work behavior — "AI auto-attends your meetings" — the way Calendly defined "send a booking link." Its competition is about who makes this behavior the default habit first, not who has better AI.

Existential threat: When Apple Intelligence or Google Gemini can natively attend your meetings, the standalone meeting product's window closes. The race is to build enough workflow embedding or habit lock-in before general agents absorb the category.

If continuing ReadyCall, the strategic question isn't "how to make better meeting notes" but "what about the meeting experience requires a dedicated product that a general agent can't do well?" Possible answers: real-time in-meeting assistance (not just post-meeting), cross-meeting organizational memory, team-level meeting intelligence.

The Meta-Reflection

After 39 turns of rigorous strategic analysis, the most important insight was the last one: sophisticated analysis can become sophisticated procrastination. Frameworks are useful for elimination (ruling out bad ideas) but not for selection. The right entry point will be found by watching someone work and noticing the moment they curse under their breath.

→ Full self-assessment in [[2026-02-06-three-startup-philosophies]]

Related

  • [[2026-02-06-what-remains-scarce]] — Main framework
  • [[2026-02-06-ai-native-saas-reinvention]] — Agent-era interfaces and SaaS reinvention
  • [[2026-02-06-three-startup-philosophies]] — Philosophy 1/2/3 deep dive and self-assessment
  • [[me.md]] — Thinking patterns captured from this session