World-Class Agentic Analytics in Production
Paras Doshi
Head of Data | Amazon, Opendoor
Build an analytics agent people will trust with real work
Founding cohort: Use code FOUNDER for 50% off, or use this link
Most analytics agents fail for boring reasons. The metric is ambiguous. The source is stale. Two dashboards disagree. The agent still gives a confident answer.
I have dealt with this as an operator. At Opendoor, my team killed 1,400 dashboards, built a semantic layer people could talk to, replaced a six-figure SaaS tool with our own AI tooling, and put agents into the daily work of every data team member.
Bring one analytics use case from your company. During the day, we will decide what the agent should do, what context it needs, how it proves each answer, and when it needs a person.
You will leave with a 30-day production plan. Every student also gets a private 60-minute session with me after the workshop. I normally charge $500 for that hour. You will also join a small community of data and AI operators doing the same work.
----------------------------------------------------------------------------
I built this at Opendoor as Head of Data, leading a global team that put AI agents into daily work. The founding cohort gets the workshop, templates, and a private 1:1 with me for $500. I will not run it at this price again.
What you’ll learn
By the end of the day, you will have a production plan for one analytics agent your team needs.
Start with a decision someone makes often. Then check the value, risk, and data needed to support it.
Write down what the agent can do, what it cannot do, and who owns the final call.
Map the metric definitions, source tables, dashboard queries, docs, and business rules it needs.
Decide when it can answer, ask a follow-up, or stop.
Require the source, query, freshness, assumptions, and caveats behind each answer.
Choose which checks run automatically and which answers need human review.
Build a small eval set from real questions, including ambiguous asks and conflicting sources.
Set a clear pass bar before anyone relies on the agent at work.
Assign owners for metrics, source changes, failed answers, and cost.
Choose a feedback loop and review schedule your team can maintain.
Turn the use case into weekly milestones, named owners, and a production checklist.
Review the plan with me in your private 60-minute session after the workshop.
Workshop agenda
Why analytics agents fail after the demo
We will look at failures that show up with real users and company data. You will choose the use case you want to fix.
Write the decision contract
Name the user, the decision, and what happens after the answer. Set the line the agent cannot cross without human review.
Build the context map
List the metrics, sources, business rules, and institutional knowledge the agent needs. Decide how it chooses among them.
Design proof and review
Make the agent show its work. Set the checks that run every time and the conditions that send an answer to a person.
Build the eval set
Turn real questions and past mistakes into tests. Define the score the system must hit before your team relies on it.
Finish your 30-day plan
Set the first milestone, name the owners, and get feedback from the group. Bring the finished plan to our private follow-up.
Who this workshop is for
You lead a data or analytics team and need to decide where an agent can do real work safely.
You are building an analytics agent and keep running into missing context, weak evals, or trust problems.
You own an AI or analytics program and need a plan your team can start in the next 30 days.
What's included
Live sessions
Learn directly from Paras Doshi in a real-time, interactive format.
Private 60-minute review with Paras
Every student gets a one-on-one session after the workshop. Bring your plan and the hardest open questions. I normally charge $500 for this hour.
Small community of data and AI operators
The cohort space stays open after class. Share what you are building, compare notes, and ask the group for help.
Your 30-day production plan
Leave with the use case, owners, weekly milestones, checks, and launch bar written down.
Templates you can use at work
Keep the context map, decision contract, eval sheet, and production checklist.
Peer review during the workshop
Get feedback on your plan from Paras and other data leaders before the day ends.
Recording and lifetime access
Rewatch the session and keep the materials.
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Frequently asked questions
Maven for Teams
Reimbursement
Get your company to pay
Everything L&D needs: email template, receipts, and certificate of completion.
Get reimbursedTeam discount
Learn with your teammates
Save 20%+ when 2 or more teammates enroll in the same cohort.
Save 20%+ with a teamPrivate cohort
Run a cohort for your org
A dedicated cohort with a custom schedule and curriculum, tailored to your team.
Book a private cohort$1,000
USD
12–5pm EDT