Define and De-Risk an AI Feature With 6 Weeks, 3 Engineers, and No ML Team
Meridian Analytics is a B2B SaaS company that sells a business intelligence platform used by operations and finance teams at 250+ mid-market companies. Their core product — a reporting and dashboard tool — has strong retention but faces…
Create a free account to upload your work. Your progress saves as a draft until you submit.
What You'll Be Doing
Meridian Analytics is a B2B SaaS company that sells a business intelligence platform used by operations and finance teams at 250+ mid-market companies. Their core product — a reporting and dashboard tool — has strong retention but faces increasing competitive pressure from AI-native entrants that offer automated insight generation.
The Chief Product Officer has committed to enterprise customers that Meridian will ship an AI Assistant feature in Q2 — a conversational interface that lets users query their data in plain language and receive AI-generated summaries and recommendations. The feature has been sold into four enterprise contracts. Failing to ship it on time risks those renewals.
You are the PM assigned to this initiative. You are 3 weeks into the role. Here is what you have learned so far:
-
Meridian has no internal ML engineering team. All AI capabilities must be delivered via third-party LLM API. The one ML engineer on staff manages existing data pipelines — she is not available for model work.
-
Four of Meridian's largest enterprise customers have contractual data residency requirements. Their data and queries cannot be sent to external APIs without explicit per-account consent and a documented data handling agreement. Legal has not yet signed off on a standard agreement.
-
Engineering capacity for this initiative is three full-stack engineers. They have six weeks before the Q2 deadline.
-
The Customer Success team has flagged that customers are cautious about AI after a competitor's high-profile AI feature failure last year (incorrect financial summaries caused two customer incidents). Trust is a real and present concern — not a theoretical one.
-
The CPO expects a written brief and roadmap before the engineering kickoff meeting in four days.
Your job is to produce the brief and roadmap the CPO is asking for — and to show your thinking about the hardest decisions you had to make along the way.
How Your Work Will Be Scored
What to Submit
AI Product Brief
Format: .pdf, .doc, .docx, .rtf, .txt, .md
-
Problem framing — what specifically the AI Assistant needs to accomplish for users and for the business in Q2
-
Feature scope — what is in Phase 1 and what is explicitly out of scope (with reasoning)
-
Architecture recommendation — your stance on how AI is delivered given the constraints (be specific about the trade-offs you are accepting)
-
A 3-milestone roadmap from now to Q2 ship, with sequencing rationale
-
2–3 success metrics — outcome-based, not output-based
-
Sign in to upload files
README
Format: .pdf, .doc, .docx, .rtf, .txt, .md
A structured written document with three sections:
Section A — Technical Constraint Assessment & Risk Register
-
Identify the top 3 technical constraints in this scenario and the specific risk each creates for the initiative
-
For each risk, state your mitigation approach — not 'we'll monitor it,' but what specifically changes in the design or timeline because of this constraint
-
State the one constraint you believe is most likely to cause a phase slip and explain why
Section B — Stakeholder Communication Plan
Write a brief (3–5 sentences each) communication for each of the following audiences. Same underlying situation — different framing:
-
The engineering team at kickoff: what they need to understand about the constraints and what you need from them
-
An enterprise customer asking when the AI feature will work with their data: honest, accurate, trust-preserving
-
The CS team who are fielding customer anxiety about AI accuracy: how to talk about the safeguards you have built in
Section C — AI Usage Log
-
List the AI tools you used
-
For each tool, describe what you prompted it to do and what you used the output for
-
Identify at least one moment where AI output was wrong, incomplete, or needed significant correction — describe what you changed and why
-
Estimate what percentage of the final submission is your original thinking vs. AI-assisted content
Sign in to upload files
Video Walkthrough
Format: .mp4, .mov, .webm
Record as an MP4 or MOV file and upload it directly on the Provn platform as a separate file.
Structure your video around these three prompts:
-
Walk us through an overview of your strategy in addition to your most important trade-off or cut decision. What did you take out of Phase 1, and what would have happened if you had not cut it?
-
How would you explain the data residency situation to an enterprise customer who wants the AI feature immediately and is frustrated that it is not available for their data yet? Walk us through what you would actually say.
-
Mandatory AI question (read verbatim — do not paraphrase): "Walk me through one specific moment in this challenge where you redirected the AI — what did it produce, why was that not quite right, and how did you correct or build on it?"
Sign in to upload files
Create a free account to upload your work. Your progress saves as a draft until you submit.
On this page