Mei Lin Yap · CTO · E-commerce · Mid-market · 110 employees
Scored from the deal record only: GoBeli is a 110-person e-commerce org whose RM105,000 proposal maps to roughly 30 workshop seats (RM3,500 each), i.e. a single cohort just under the 35 cap with the remaining ~80 staff unaddressed. The two biggest gaps are unowned tooling and a completely uncaptured baseline — no Cursor/Claude Code/Copilot licences, MCP config, IDE rules or secrets policy appear anywhere, and there is no cycle time, PR-throughput, deploy-frequency or escaped-defect number to prove the workshop changed anything. The highest-leverage 90-day move is to run Day 1 as a provisioning lab so all 30 seats leave with live licences, a shared rules repo and a written agent secrets policy, then instrument a four-week before/after scorecard so a phased cohort 2 is justified on evidence rather than enthusiasm.
The deal record names no AI editor or agent tooling at all — no Cursor, Claude Code or Copilot seats, no MCP server inventory, no .cursorrules/CLAUDE.md artefacts and no secrets-handling policy. That absence is the reason this sits at 54 rather than higher: the CTO is buying a tools-first Day 1, but nothing on file shows the stack exists to land it, so the workshop must include provisioning rather than assume it.
Nothing in the record shows AI woven into PR review, testing or code-gen: no review policy for AI-authored diffs, no agent guardrails, no prompt-hygiene standard. The score is 42 because Day 2's agentic-development and guardrails content is currently inapplicable — the prospect has no existing AI workflow to integrate into, meaning Day 2 has to build the workflow rather than refine it.
Leadership posture is the strongest signal in the file: the CTO, Mei Lin Yap, is the named contact and personally carries a proposal-stage deal, and the account carries a CRM health of 78. It stays at 68 rather than healthy because the deal is explicitly tagged 'Phased rollout', indicating deliberate, low-blast-radius sequencing rather than a mandate for autonomous agents.
No throughput baseline was captured before proposal: cycle time, PRs per developer per week, time-to-merge and deploy frequency are all blank in the record. It scores 48 because the workshop's core commercial promise — closing a measurable velocity gap — currently has no denominator; the first deliverable has to be instrumenting these four numbers.
Adoption is unmeasured, not merely low: the record contains no daily or weekly AI usage figure and no licence-utilisation number. The one concrete observation is scale — ~30 seats against 110 employees is roughly 27% of headcount, so even perfect workshop attendance leaves the majority of the org untouched unless a second cohort is phased in.
Output-quality data is entirely absent — no defect rate, escaped-bug count, time-to-feature or developer-satisfaction figure appears in the record. It scores 50 because the account is commercially healthy (deal stage Proposal, probability 60) but the quality story the workshop would be judged against is unevidenced, so impact claims risk being anecdotal.