Failure Analysis

Why We Lost a £200k Contract: An AI Automation Failure Post-Mortem

We promised results we couldn't deliver on time. This is the full post-mortem — what broke, what we missed in scoping, and the checklist we now use before every engagement.

AR

Alex Rowland

Founder & CEO · 28 March 2026 · 8 min read

Key Takeaways

An Aroluxa AI automation engagement failed to deliver within the promised timeline, resulting in contract termination and £200k revenue loss. Root causes: underestimated data cleaning requirements, undocumented client workflows, and overconfident timeline scoping. The failure produced three systemic changes: a mandatory data audit before engagement, a workflow documentation sprint as project phase 0, and conservative timeline commitments with milestone gates. This post-mortem is shared publicly to establish credibility through transparency and to help other AI agencies avoid identical scoping failures.

failure analysisAI implementationagency mistakesscopingpost-mortem

The Engagement That Went Wrong

We signed a 12-month AI automation contract with a mid-market logistics company. Scope: automate their dispatch coordination, invoice processing, and driver communication workflows. Our projection: 70% time savings within 90 days. We were wrong — not about the outcome, but about how long it would take to get there. We hit 70% savings at month 7, not month 3. They cancelled at month 5.

Root Cause #1: Data We Didn't Know About

The client had three separate CRM systems. We knew about two. The third — a legacy system used by one regional team — contained 40% of the customer data we needed. Cleaning and migrating that data took 11 weeks. Our original estimate: 2 weeks. We should have run a data audit as the first deliverable, not assumed the client's data description was complete.

Root Cause #2: Undocumented Tribal Knowledge

The dispatch coordinator role relied on institutional knowledge accumulated over 9 years. When we mapped the workflow in our pre-engagement calls, we documented the 'official' process. The actual process — the one that handled 60% of edge cases — existed only in one person's head. We discovered this when our AI agent started making decisions that the coordinator called 'obviously wrong.'

Root Cause #3: We Oversold the Timeline

We wanted the contract. We gave them a timeline we were 65% confident in, not the 90% confidence timeline we use internally. That 25% confidence gap cost us £200k and, more importantly, cost them time and trust. Every agency faces this pressure. We now have a rule: never commit to a timeline below 85% internal confidence, regardless of what the client wants to hear.

The Three Changes We Made

1. Data Audit as Phase 0 — before signing any SOW, we run a 2-week paid data audit. Non-negotiable. 2. Workflow Documentation Sprint — the client's team documents their actual workflow, not the official one, before we touch any code. 3. Conservative Milestones — we now build in 40% buffer on all AI implementation timelines. We deliver faster than promised, not slower.

Ready to implement this in your business?

Book a free AI Audit. 90 minutes. We'll map your highest-value opportunities and hand you a prioritised implementation plan.

Book My AI Audit

Related reads

Failure Analysis11 min

Why 67% of AI Agency Engagements Fail in Year One (And How to Not Be That Client)

After interviewing 34 businesses who churned from AI agencies, we found three failure patterns that account for 91% of failed engagements. Here's what they are and how to avoid them.

AI agencyfailure patternsclient success
AR

Alex Rowland

7 March 2026