Dashboards founders actually open
Seven systems I designed and built for founders and leadership teams, each with its schema and the rules a person signs off on. Numbers on screen are illustrative; the rules and the results are real.
Evidence ranks the options. A person makes the call.
Leadership was choosing between competing venture models, with the evidence spread across long discovery documents. Every recommendation had to show its sources, and no model was allowed to make the call on its own.
What I built
- Decision science dashboards that score competing venture models against 1,000 predefined success markers across five functions.
- Gate tiles per concept (pass, conditional, fail, untested) with every number carrying a source chip back to the document and section.
- An evidence-coverage score from five checks, including zero numbers on screen without a source.
- A logged stop condition moves the call back to review; the model never switches to the alternative by itself. Sign-off needs a named owner and the prior gate met.
Impact
Where this happened: A venture studio · human-in-the-loop dashboards for comparing venture models against success markers; strategy phase compressed to four days
Every founder knows where their build stands.
Founders on a 12-week venture build were getting updates by message thread. They needed one place that shows the phase, the next gate, the decisions waiting on them and nothing else.
What I built
- A client delivery view across a 12-week build and five phase gates, with stages derived from events, never set by hand.
- An engagement health score weighted across responsiveness (0.30), delivery (0.25), product (0.20), commercial (0.15) and cadence (0.10).
- A weekly report drawn only from internal reviews that were approved, so nothing unreviewed reaches the founder.
- A shared decision log and handoff system across seven project boards so studio, client engineering and partners read one source of truth.
Impact
Where this happened: A venture studio · client delivery dashboard, weekly report, shared decision log across seven boards
One graph for what the firm knows and why it decided.
Research, client work and team expertise lived across documents, boards and people. Leadership needed contextual search and recommendations that cite their source, and decisions that anyone could trace.
What I built
- A firm-wide knowledge system with structured intake and a graph linking team members, client projects, initiatives and source records.
- Decision governance as part of the shared context: decision records with context, options, reasoning, reversibility and what they supersede.
- One home per type of fact. Other planes link by pointer and never copy, and a drift check flags when they diverge.
- Decision rights by type, so posture, commercial, architecture and implementation calls each have a clear owner.
Impact
Where this happened: A venture studio · firm-wide AI knowledge system and organizational knowledge graph; operating playbook
Reps work the accounts that are actually in market.
A lean, founder-led sales team was working flat lists. They needed accounts ranked by fit and timing, with the evidence behind each score and a first message ready to review.
What I built
- An intent engine that reads public buying signals (hiring, funding filings, press, community requests) and scores them in log-odds so volume can never beat quality.
- ICP fit and intent kept as separate axes, A to D qualification, and a measured AI-visibility gap that routes accounts to the right play.
- Fourteen gates that fail closed, including lawful basis by region, cite-everything, tenant isolation and a rule that scores recompute only from evidence actually collected.
- Openers drafted from cited evidence only, approved by a person before anything is sent. Demand engines on Clay, Apollo and Smartlead behind it.
Impact
Where this happened: Tenpoint Labs · agentic ICP, signal scoring, outbound; Anthony Venture Labs · intent systems; working paper on costly public signals
A remote team, measured on proof, not presence.
Virtual assistants were supporting two businesses at once: an AI FinOps product for construction firms and a studio that builds sites and MVPs for trades. The founder needed attainment, hours and proof of work in one view, and work that fed the pipeline directly.
What I built
- Role-based workspaces for founder, manager and VA, with KPIs on daily, weekly or monthly cadence and a weighted performance score.
- A review queue: every work log is approved or flagged with a reason, with proof uploads and a dated performance report.
- An opportunity engine scored as 100 × fit^0.4 × intent^0.6, sorted into close, chase, farm or ignore, with one VA per lead and a cap of 15 active picks.
- An autonomous agent that sources leads every three hours from open data, verifies them adversarially and leaves outreach as drafts.
Impact
Where this happened: Keystone AI and Anthony Venture Labs · founder operating system; SOP and SLA structures; KPI design
The whole company in a 60-second read.
Founders were running a multi-country delivery portfolio from status meetings. They needed financial, operational and project health in one place, SLAs that hold, and the repeat work taken off people.
What I built
- A Today view that ranks everything by severity: breached blockers first, then P0 and P1, drifting key results, missing KPI data and missed standups.
- A health score per unit from growth, retention, runway and execution (30/25/25/20), and key-result drift against a straight line across the quarter.
- Blockers scored impact × urgency into P0 to P3, each with an SLA (1, 3, 7, 14 days), cycle time, first-pass acceptance and automation ROI tracked.
- A copilot that drafts the change; a person confirms it. No resolve without a note, no deletes, and any task that recurred three times got automated.
Impact
Where this happened: RISE Consulting · internal operating system, SOPs and SLAs; Riseapp.ai · 25% fewer delays; a venture studio · cycle time, first-pass acceptance, automation ROI
Every piece of feedback ends in a verified change or a reason.
A health AI startup was building a reactive assistant. Interviews, pilot calls and client reviews held the answer, but feedback was scattered and nobody could say what had changed because of it.
What I built
- A loop from captured to tagged, decided, shipped and verified, with every item carrying its source and who said it.
- An interview scorecard of five qualification signals; four or more advancing qualifies, and a hard miss on data or sponsor disqualifies.
- Pilot feedback from enterprise call centers turned into fixes that moved an AI agent toward production readiness.
- An adversarial review pass in a fresh context before anything ships; it once caught 26 findings that self-review missed.
Impact
Where this happened: Riseapp.ai · discovery and pivot, activation and lead-quality metrics; RISE · Resolve AI pilot feedback; Commit Fellowship · customer discovery