Home· Knowledge· Toolkit· Agent Roadmap Template
Toolkit · Notion · Free
🗺

Agent
Roadmap Template.

90 days, structured. From the first audit to a production multi-agent setup. This Notion template contains every task, milestone, ownership assignment and review point we use internally for every client project.

📄 Notion Template 📅 90-Day Plan ✅ 47 Tasks
Duplicate Template → View Timeline
Contents

What's included in the template.

📅
3-Phase Timeline
90 days split into 3 phases: Audit (Day 1–14), First Setup (Day 15–60), Scaling (Day 61–90).
✅
47 Predefined Tasks
Every task with a description, estimated time, dependent tasks and a suggested owner.
👥
Ownership Matrix
RACI matrix for 8 roles: Sponsor, Owner, Implementer, Tester, Compliance, Stakeholder.
🎯
7 Milestones
Clearly defined go/no-go points. Every milestone with acceptance criteria — no vague success claims.
📊
Risk Register
12 typical risks with probability, impact and mitigation strategy. Pre-filled from 150+ setups.
📝
Status Reports
Templates for weekly status updates to stakeholders. One A4 page, three status colors, clear asks.
Timeline

90 Days in 3 Phases.

The timeline isn't set in stone — some setups go into production after 60 days, others need 120. But the order of the phases stays the same, and any attempt to skip one leads right back to the start.

Day 1–7 · Audit & Use-Case Selection

Stock-take: where do the hours go? Which processes are candidates? Output by the end of the week: three prioritized agent use cases with an ROI estimate.

Build the process inventory (Sheet 1 from the Audit Template)
Run the scoring matrix on the top 15 processes
Stakeholder interviews (Owner, Sponsor, Compliance)
Document the top 3 candidates with reasoning
Go/no-go meeting at the end of week 1

Day 8–14 · Setup Preparation

From use case to a concrete plan. Which tools? Which data sources? Which escalation paths? Output: an implementation doc with every architecture decision.

Decide the tool stack (n8n, Make, custom — depending on complexity)
Clarify data access (API keys, permissions, OAuth)
Work through the GDPR checklist (separate toolkit file)
Define success metrics — 2 to 3 KPIs, no more
Prepare the sandbox setup (separate data, clearly bounded)

Day 15–30 · Sandbox Phase

The agent is running, but NOT on real customer data. Synthetic or historical data. Iteration in short loops. Output: a stable V1 that correctly resolves 80%+ of test cases.

First prompt version, 20 historical test cases
Prompt iterations — 2-3 refinements per day
Edge-case collection in a separate database
Build and test escalation logic
Daily stand-ups with the owner — what worked, what broke

Day 31–45 · Supervised Live Operation

The agent goes live, but with active supervision. Every output is reviewed before it goes out. Goal: build trust — from both the human and the stakeholders.

First 100 live outputs with human review
Weekly review round with the owner
Error tracking in the bug database
Re-training after 50 live cases
Stakeholder update at the end of week 6

Day 46–60 · Trust Handover

Reviews become less frequent. Auto-approval for clear categories, escalation only when there's uncertainty. Output by the end: the agent runs autonomously in 80%+ of cases.

Define thresholds for auto-approval
Monitoring dashboard live (KPIs in real time)
Escalation workflow for edge cases documented
Owner handover doc written
Milestone review: Agent #1 in production

Day 61–75 · Stabilization & Scaling

The first agent is running. Don't build a second agent now — harden the first one instead. Iterate on edge cases, performance, cost. Output: a robust V1 with a 6+ month lifespan.

Edge-case sweep: what's come up in live operation?
Cost optimization (model choice, token efficiency)
Performance tuning (latency, throughput)
Backup strategy and failover paths
First KPI review: ROI against expectations

Day 76–90 · Second Agent & Roadmap

Only now — once Agent 1 is stable — does work on Agent 2 begin. With the learnings from Setup 1, it goes twice as fast. Output: a 2-agent stack + a multi-quarter roadmap.

Use-case audit for Agent #2 (3 days)
Setup phase for Agent #2 (shortened by learning effects)
Memory-layer concept if multi-agent is planned
Roadmap for Agents #3–5 for the next 6 months
Quarterly review with all stakeholders
The Most Important Rule of the Template

Don't skip a phase — no matter how tempting. We've seen it across 150+ setups: skip the audit, and you fail in phase 3. Skip the sandbox, and you fail at the first live edge case. Skip stabilization, and in 6 months you have two half-finished agents instead of one whole one. The order is not a suggestion.

Ownership

Who does what?

Included in the template: an ownership matrix with eight roles. Here's a shortened overview of the most important ones:

★
Sponsor — usually the CEO/owner
Carries the initiative politically. Clears obstacles out of the way. Makes the go/no-go decision. Not involved in the detail, but in the loop for every milestone.
▶
Owner — operational responsibility
Involved daily. Drives progress. Escalates on blockers. This role is the single most important success factor — without a clear owner, every setup dies.
⚙
Implementer — builds the agent
Internal or external. Responsible for technical implementation. Reports to the owner, not the sponsor.
✓
Tester — domain experts from the team
Reviews outputs during the sandbox phase. Important: they are not the implementers. Fresh eyes catch different mistakes.
⚖
Compliance — for GDPR-relevant setups
Data protection officer or external advisor. Involve early — bringing them in at the end is always more expensive.

Use the Template Right Away.

Click the button — the Notion template is duplicated into your workspace. Including all 47 tasks, the ownership matrix and the risk register templates.

Duplicate Template →