Summit Advisory Group is a 600-person management consulting firm specializing in operational transformation for mid-market companies. On a Monday in February, the CEO sent an all-hands email announcing that Summit would become "an AI-first firm" and would "deploy AI across all client delivery and internal operations workflows within 90 days." You were named the AI Transformation Lead in the same email. The email was well-intentioned, enthusiastically received by some employees and anxiously received by others, and almost entirely lacking in specifics. You have no pre-existing governance framework, no vendor contracts beyond individual employees' personal subscriptions to AI tools, no approved AI policies, and no dedicated budget. The board has been briefed at a high level and is expecting a transformation plan at the next quarterly meeting in 10 weeks. Design the 90-day plan.
The first mistake a transformation lead can make in this situation is treating the CEO's announcement as a project specification. It is not. It is an organizational signal — a statement of strategic intent that creates authority and urgency for a transformation program that still needs to be designed. The 90-day timeline is aspirational, not operational; no serious governance framework, no rigorous workflow selection, and no meaningful training program can be completed in 90 days if started from zero. Your job in the first two weeks is not to execute the announcement — it is to replace the announcement with a credible plan that achieves the CEO's intent while being honest about what can and cannot happen in 90 days. This requires a direct conversation with the CEO that most transformation leads avoid because it feels like pushback on a strategic mandate. It is not pushback. It is the professional responsibility of someone who understands what transformation actually requires.
The 90-day plan has four phases. Days 1–15: Discovery and Triage. Inventory existing AI tool usage across the firm — who is using what, for what purpose, with what data. Identify the five workflows with the highest pain and clearest measurable outcomes. Classify each for risk. Conduct a stakeholder survey to understand both enthusiasm and concern. Establish a temporary governance working group with representatives from legal, technology, HR, and client delivery. Days 16–45: Foundation and First Pilot. Draft the core governance policy: approved tools, allowed data classes, prohibited uses, and incident reporting procedure. Select one pilot workflow from the five candidates — the one with the best fit score, lowest risk classification, and a clear owner who is genuinely motivated. Launch the pilot with explicit acceptance criteria, a human review path, and weekly reporting to the CEO. Begin firm-wide AI literacy training for all staff. Days 46–75: Pilot Review and Expansion Plan. Conduct a structured retrospective on the pilot at the six-week mark. Present findings to the CEO and board: what worked, what did not, what the economics show, and what the revised expansion plan looks like. Use this meeting to replace the "90 days, all workflows" narrative with a "phased, governed, evidence-based expansion" narrative — framed as the more rigorous version of the original mandate, not a retreat from it. Days 76–90: Operational Model and Roadmap. Document the governance framework, the workflow selection criteria, the pilot learnings, and the 12-month expansion roadmap. Present to the board. This is the deliverable that the board meeting requires — not 600 employees using AI, but a credible plan for how the firm will develop AI capability responsibly over the next year.
The success metrics for this plan should be defined before execution begins, not after. At 90 days: one pilot workflow in production with documented performance against acceptance criteria, governance policy adopted and trained across all staff, five additional workflows in the pipeline with risk classifications and owners, and board approval of the 12-month roadmap. At 12 months: three workflows in stable production operation, ROI documented for at least two, governance framework reviewed and updated based on learnings, and a named AI capability function with defined roles and a budget. These metrics are honest, achievable, and meaningful. They are also significantly more modest than what the CEO's original announcement implied — and explaining the gap between the announcement and the plan is the first and most important communication challenge of the transformation lead's tenure.
What this means in practice
The practical implementation question is not whether the idea is interesting. It is how a team turns it into a workflow that can be inspected, repeated, and improved. For this topic, the operating focus is direct: Practice the transformation lead's first responsibility: replacing a strategic announcement with a credible plan — and building the 90-day roadmap that achieves the CEO's intent while being honest about what can and cannot happen.
That means the engineering work starts before the first model call. The team must decide what the agent is allowed to know, what it is allowed to do, what evidence it must produce, and which actions require a human decision. This is the difference between an impressive demo and a system that can survive real users, changing inputs, and production constraints.
A credible implementation also includes a feedback path. Every agent run should leave behind enough context for another engineer to answer four questions: what goal was attempted, what context was used, which tools were called, and why the system believed the task was complete. If those questions cannot be answered from logs, traces, or structured outputs, the agent is still operating as a black box.
A simple architecture to reason from
Use this diagram as a starting point, not as a universal blueprint. The important move is to make the stages visible. Once stages are visible, you can assign owners, define contracts, set permissions, measure quality, and decide where human review belongs.
'AI-first in 90 days' = organizational intent, not a project specification.
Replace the announcement with a credible plan — this is professional responsibility, not pushback.
Inventory AI usage, identify top 5 workflows, classify for risk, form governance working group.
Draft governance policy, select pilot, establish acceptance criteria, begin AI literacy training.
6-week retrospective: present findings, replace '90 days all workflows' narrative with phased expansion.
Document governance framework, pilot learnings, and 12-month roadmap.
Deliverable: credible plan + one functioning pilot — not organization-wide deployment.
90-day transformation plan structure
The example below is intentionally small. Production agentic systems should start with compact contracts like this because small contracts are testable. Once the boundary is working, you can add richer orchestration without losing control of the core behavior.
const transformation90DayPlan = {
phase1_discovery: {
days: "1-15",
deliverables: [
"Inventory of existing AI tool usage across the organization",
"Top 5 workflow candidates with pain and outcome assessment",
"Risk classification for each candidate",
"Stakeholder survey: enthusiasm and concern landscape",
"Governance working group formed (legal, technology, HR, client delivery)",
],
},
phase2_foundation: {
days: "16-45",
deliverables: [
"Core governance policy: approved tools, allowed data, prohibited uses, incident reporting",
"Pilot selected: best fit score + lowest risk + motivated owner",
"Pilot acceptance criteria defined and reviewed",
"AI literacy training launched for all staff",
],
},
phase3_review: {
days: "46-75",
deliverables: [
"6-week pilot retrospective: what worked, what didn't, economics",
"CEO and board presentation: findings + revised narrative",
"Replace 'all workflows in 90 days' with 'phased, governed, evidence-based expansion'",
],
},
phase4_roadmap: {
days: "76-90",
deliverables: [
"Governance framework documented",
"12-month expansion roadmap",
"Board presentation: credible plan + pilot results",
],
},
successMetrics: {
at90Days: [
"One pilot in production with documented performance against acceptance criteria",
"Governance policy adopted and trained across all staff",
"Five workflows in pipeline with risk classifications and owners",
"Board approval of 12-month roadmap",
],
},
};Implementation notes
Treat these notes as the first design review checklist. They are deliberately concrete because agentic systems fail most often in the gaps between the model, the tools, the data, and the human operating process.
The CEO's transformation announcement creates authority and urgency — not project scope. Replace it with a credible plan.
The first and most important communication challenge is explaining the gap between the announcement and the plan — frame the plan as the more rigorous version of the mandate.
Define success metrics before execution begins — at 90 days and at 12 months — so the board presentation has a measurement basis.
Common failure modes
The fastest way to make an article useful is to name how the pattern breaks. These are the failure modes to watch for when a team moves from reading about this idea to deploying it inside a real workflow.
Operating checklist
Before this pattern graduates from experiment to production, require a short operating checklist. The checklist should include the owner of the workflow, the allowed tools, the risk rating for each tool, the data sources the agent can use, the completion criteria, the review path, and the rollback plan. If a team cannot fill out that checklist, the workflow is not ready for higher autonomy.
The checklist should also define how the system will be evaluated after launch. Useful metrics include task success rate, human correction rate, average iterations per completed task, cost per successful run, escalation rate, and the number of blocked tool calls. These metrics turn agent quality into an engineering conversation instead of an opinion about whether the output felt good.
Finally, make the learning loop explicit. When the agent fails, decide whether the fix belongs in the prompt, the retrieval layer, the tool contract, the permission model, the evaluation suite, or the human process. Mature agentic engineering is not the absence of failures. It is the ability to classify failures quickly and improve the system without expanding risk.
Build real fluency in agentic engineering.
The Academy turns these concepts into a full curriculum, AI tutor, templates, and the CAE credential path.