Dell’s Amy Mitchell on Why AI Is Creating the “Builder PM”
AI is transforming product management faster than most teams realize. Dell’s Amy Mitchell explains why the next generation of PMs won’t just define requirements, they’ll step in and build.
AI Evolution: AI transforms product managers into 'builder PMs' who tackle obstacles and enhance collaboration early in projects.
Continuous Improvement: Utilizing AI tools enables ongoing optimization of workflows, enhancing delivery speed and team efficiency.
Knowledge Management: Integrating Jira and Confluence allows teams to self-serve, reducing administrative burdens and increasing focus on strategy.
Agile Methodology: Transitioning to hybrid agile simplifies processes, relying on fewer tools for effective reviews and streamlined communication.
Delivery Speed: Accelerated work processes lead to more frequent releases, with straightforward tasks increasingly managed by product managers.
Amy Mitchell is Principal Product Manager at Dell, where she focuses on bringing managed services offers from concept to delivery. With experience across networking, storage, and data center infrastructure, she works closely with customers and delivery teams to help organizations modernize and scale through managed services. She's also the creator of the Product Management IRL newsletter.
We caught up with her to get a sense of how the discipline is changing, and what that means in practice. Here's what she had to say.
Delivering managed services
I'm one of those product managers who is interested in how my products reach customers. In my current role as Principal Product Manager at Dell, I get to work closely with program managers and service delivery teams in launching and maintaining managed services.
Unlock for Free
Create a free account to finish this piece and join a community of forward-thinking leaders unlocking tools, playbooks, and insights for thriving in the age of AI.
Have an account? Log In
The "products" I deliver aren't hardware or software products. My products are managed services like managed storage, managed AI factory, or managed compute, to name a few.
Outside of that work, I also write the Product Management IRL newsletter.
Why AI is forcing PMs to become "builder PMs"
With AI, my role is evolving into what I call a "builder PM" role. A "builder PM" is a product manager who steps out of their lane to unblock work.
With AI, my role is evolving to a “builder PM” role. A builder PM is a product manager who steps out of their lane to unblock work.
When I'm in builder-PM mode, I work closely with the product team early in the development cycle — often working with engineering, sales, and marketing before writing a single requirement.
I can do this because AI augments my work and allows me to zoom out. Here are a few examples of AI's value adds in my line of work:
- Market and pricing research: This allows me to quantify use cases for pricing and forecasting.
- Detailed roadmaps that verge on a program plan: Speedy drawings and critical path mapping.
- Collaborative workspaces: Automated updates and self-serve information to the whole product team.
- Discernment of complex situations: Perspectives on conflicts and resolutions.
- Meeting minutes: Automated minutes free up a program manager's time every day.
- Vibe coding: Scripts to automate documentation updates.
How AI supports continuous improvement in product and delivery workflows
The value doesn't stop there. Where I work, we believe in continuous improvement — and we use AI heavily for this.
Homegrown AI helps us with the post-mortems that are used to identify workflow improvements. Our internal copilot assists with researching ways to optimize the workflows with customers. And standardizing how we use AI in development has enabled automation via APIs and scripts.
But some things still require a human touch. Sales contact and customer conversations, for example. And solutioning of new capabilities.
How AI-enhanced knowledge bases free up PM time
By integrating our work management platform — Jira — with a centralized knowledge repository — Confluence + SharePoint — we’ve been able to capture complex workflows and service descriptions in a way that allows the delivery team to self-serve, significantly reducing the administrative load on our program managers.
In other words, the product team can be consistent and find what they need without a meeting or IM interruption. And PMs can spend more time anticipating blockers to customer delivery.
Our product portfolio's knowledge base is a group of shared folders. The latest versions are updated monthly and posted on Confluence. The whole product team contributes to the monthly version update and pulls what they need to post on sales pages, announcements, meetings, or use in daily work.
The major items in the knowledge base are:
- Release notes on what changed
- Ordering guide
- Contract templates
- Configuration and pricing calculator
- Sales training
- FAQ
Jira serves as the orchestration platform that interacts with the knowledge base. Up to this point, my effort was on creating the knowledge base and workflows. Automation through AI assistance is next!
How to roll up status from Jira and Confluence without slide decks
We also recently started using Confluence to roll up status from our PRD (also in Confluence) and Jira functional plans.
Instead of multiple meetings refining a slide deck before executive gate reviews, we now read out the plan and the launch summary to executives straight from Confluence.
And we use this same material for weekly program tracking.
How evals and orchestration are becoming core PM skills
I've borrowed two key AI methodologies for use in product management: evals and orchestration.
At their core, evals are systematic, repeatable ways to validate outcomes. They first gained traction in AI, where outputs change constantly and drift is real, but the concept is much broader. I believe evals are becoming a core skill for PMs.
The beauty of evals is speed. Instead of exhaustively testing every permutation, you check a small, prioritized set of customer-centered scenarios that reflect what really matters. Here's the flow:
- Prioritize use cases.
- Set up the environment — just enough to test critical paths.
- Evaluate and iterate accordingly.
Orchestration isn't just for AI anymore, either. It's how you connect strategy and delivery, solving the risk of the handoff. And it's a necessary skill for PMs now, with organizations becoming flatter and delivery moving faster than ever.
Orchestration isn’t just for AI anymore, either. It’s how you connect strategy and delivery, solving the risk of the handoff. And it’s a necessary skill for PMs now, with organizations becoming flatter and delivery moving faster than ever.
In this context, orchestration means:
- Mapping dependencies
- Building lightweight orchestration rituals. These vary a lot from team to team. Currently, we review requirements in the product team and get agreement before assigning stories out. When we are adding a new capability, we do "solutioning" to outline the major workflows on a white board.
- Making risks visible early
When you do this across teams, it prevents gaps, earns leadership trust, and accelerates your career.
What an AI-first PM tool stack looks like
It's important to learn how to use a few good AI tools! They'll save you a huge amount of time.
But don't waste your time on tools that IT doesn't endorse. You don't want to open your product to security, privacy, or ethics issues by using non-sanctioned AI tools.
Don’t waste your time on tools that IT doesn’t endorse. You don’t want to open your product to security, privacy, or ethics issues by using non-sanctioned AI tools.
As I mentioned, we use Confluence for our PRDs and knowledge-base management, and Jira for delivery. Both have AI integrations.
I also rely heavily on Microsoft Copilot for research and summarization. And we have homegrown AI tools, like our internal copilot and support tools.
For evals, I'm big on simple spreadsheets. They save time. We recently added a new CPQ (configure, price, quote) automation tool that generates complex quotes and contracts in minutes instead of days. We completed the testing of the new tool in two weeks after taking a month to load our product data. We had five evaluators and twelve use cases. We cycled through five times with fixes on each cycle. We used a spreadsheet to track our async work and report progress to stakeholders, and it was highly effective.
And personally, I leverage ChatGPT, Gemini, Copilot, Perplexity, Visual Studio, and GitHub.
How Confluence and Jira enable hybrid agile without heavy gate reviews
As far as project management methodologies, we used agile in the past, with heavy gate reviews for each step of the program lifecycle. The tools that we used were Word docs, PowerPoints, and some Jira.
Now, we use hybrid agile with only a plan review and a release review. The only tools we need for this are Confluence and Jira. Like I said, we put everything product teams, stakeholders, and execs needed into Confluence. One place, self-serve. That alone changed our methodology.
Why delivery speed is accelerating and what it means for PM roles
I see two big changes coming.
First, the speed of everyone's work is accelerating daily. Just with these tools mentioned above, I've gone from one or two releases per year to monthly releases — and that change took place in a matter of months.
That's not going to stop. New tools — and PMs getting better at using these tools — will continue speeding up product development.
Second, I think program managers will only handle the most complex and technical programs. Straightforward changes will be made by product managers — and the tools themselves.
How to avoid AI theater and build what can actually ship
My advice is to get creative. Focus on the best customer outcomes. And quit doing impressive-looking demos that can't actually be launched. That's the line between AI-first product management and AI theater.
Quit doing impressive-looking demos that can’t actually be launched. That’s the line between AI-first product management and AI theater.
Too many PMs have built impressive AI demos that look right and move fast. But then, they realize the solution cannot be shipped.
It's not about demo quality, it's about outcome alignment.
Follow along
You can follow along as Amy Mitchell shares her learnings about real-life product and program management on the Product Management IRL newsletter. Or follow her on LinkedIn.
More expert interviews to come on The Digital Project Manager!
