Overview of #203: Every Software Team Should Put AI Agents on the Org Chart - Nick Olsen
In this episode of the Practical Founders Podcast, Greg Head talks with Nick Olsen, VP of AI Innovation at Mainsail Partners, about how AI is reshaping product and engineering teams inside bootstrapped and growth-stage vertical SaaS companies. Nick argues that the industry has moved past skepticism: AI is no longer just a productivity tool, but a core part of how modern software teams should operate. The conversation focuses on practical, hands-on ways to adopt AI in development workflows, product management, org design, and customer value creation.
Who Nick Olsen Is and What Mainsail Does
Nick shares his path from computer science and founding/operator roles to leading AI innovation across Mainsail’s portfolio.
- He co-founded and scaled Resmond, a property-management software company for multifamily assets.
- He later joined Mainsail Partners, a growth equity firm focused on vertical SaaS.
- At Mainsail, he leads a team of five AI engineers that works directly with portfolio companies.
- Mainsail’s “growth team” is hands-on and operational, not just advisory or capital-focused.
A key point: Mainsail doesn’t just fund companies—it embeds operators who help them build and scale better systems, including AI capabilities.
The New Baseline for Engineering Teams
Nick’s core message is that the baseline for software teams has changed dramatically.
AI is no longer optional
- Teams are past the point where AI can be dismissed as temporary or hype.
- Engineers who still refuse to use AI are increasingly seen as out of step with the market.
Engineers should start with agents, not IDEs
- The traditional code editor is no longer the default starting point.
- Development is shifting toward an agentic workflow using tools like:
- Claude Code
- Codex
- The goal is not just faster coding, but a different way of working altogether.
AI should be treated as a teammate
- Nick says companies are moving from “AI as a tool” to “AI as part of how we operate.”
- The strongest teams treat AI like a teammate with real responsibilities.
How High-Performing Teams Are Structuring AI Work
Mainsail’s approach is deliberately hands-on and embedded.
The operating model
- Mainsail forms a pod with:
- one or two engineers from the portfolio company,
- a product manager,
- and one or two engineers from Nick’s team.
- They commit code directly into the company’s codebase.
- The engagement is temporary but the relationship continues long-term.
Why this matters
Nick strongly pushes back on the common consulting model of:
“We’ll send in a team, make you AI-first, and report back later.”
Instead, he argues for side-by-side implementation, coaching, and code-level collaboration.
What Changes in Product and Engineering
Nick says the real shift is not just in tooling, but in team structure and process.
Handoffs are disappearing
Traditional workflow:
- product gathers requirements,
- engineering builds,
- QA checks,
- then release.
AI-enabled workflow:
- product, design, and engineering work together earlier,
- prototypes are created quickly,
- the team stays closer to the customer problem.
Roles are collapsing
He sees product managers, engineers, and designers becoming more fluid in practice:
- Product managers are getting more technical.
- Engineers are becoming more product-minded.
- Teams are working from shared context earlier in the process.
Product discovery can be AI-assisted
Agents can help product teams:
- analyze backlog items,
- review support tickets,
- scan call recordings,
- extract customer patterns,
- support prioritization with more evidence.
This helps teams make better decisions with more context than any one person can hold.
Measuring Teams by Outcomes, Not Output
A major theme in the conversation is the move away from traditional engineering metrics.
Old model
- Story points
- PR counts
- Velocity
- Output-based evaluation
New model
- Customer impact
- Business outcome
- Time to value
- Meaningful adoption
Nick argues that AI shortens cycle time so much that old metrics become less useful. Teams should be measured by whether they move a customer or business result, not how many tasks they completed.
Putting Agents on the Org Chart
One of the most memorable ideas in the episode is Nick’s recommendation that every company should literally put agents on the org chart.
Why do this?
- It creates accountability.
- It forces the organization to define the agent’s job clearly.
- It makes the agent something the team can evaluate like any other teammate.
What that means in practice
If an agent isn’t delivering value:
- you don’t just discard it,
- you improve its instructions, data, and harnesses,
- you give it a better-defined job.
This turns agents into operational assets rather than novelty tools.
Codifying Internal Expertise into Agents
Nick explains that the most valuable agents are not generic off-the-shelf tools.
Best use case
- Identify the people in your org who know the “brittle” or high-risk areas best.
- Extract and codify their expertise into agents.
- Run those agents continuously across the workflow.
Examples:
- a code review agent trained on your most trusted reviewer’s standards,
- a security agent that knows your company’s special risk areas,
- a testing agent that understands business-specific edge cases.
This is how AI becomes leverage, not just automation.
The Right Way to Roll Out AI in a Team
Nick says successful adoption requires time, space, and retraining.
Key advice
- Don’t just hand everyone a license and hope for the best.
- Give teams time to slow down and learn new workflows.
- Use focused implementation windows to build harnesses and skills.
- Expect some resistance from senior leaders and engineers.
Leadership must adapt
He says the best CTOs and heads of engineering are now:
- hands-on,
- back in the code,
- willing to learn alongside their teams.
He also suggests using a 1-to-5 skill ladder for AI adoption:
- beginner
- intermediate
- adopter
- evangelist
- expert
This makes expectations explicit and gives teams a path forward.
Legacy Code, Rewrites, and the Future Platform
Nick is more open than many leaders to the idea of rewriting systems if needed.
His view
- The “rewrite” is more feasible now than it was a couple of years ago.
- But it should start with a question about the future platform, not the current one.
- Ask:
- What will customers need in 2027?
- What must the platform support then?
- Can the current architecture support it?
If not, a rewrite or major re-architecture may be justified.
AI changes what’s possible
He believes that with the right team, tooling, and SME input, many software platforms could be rebuilt in six to eight months.
Product Moat Is Not Dead, But It Is Expanding
Nick rejects the idea that AI makes product irrelevant.
His argument
SaaS companies still win through:
- deep vertical expertise,
- embedded workflows,
- proprietary data,
- customer education,
- ecosystem value,
- and business logic.
But the moat is broader now:
- It’s not just the app.
- It’s the software plus the knowledge, guidance, and operational support around it.
He believes successful software companies will become more like platform + education + workflow partner than just a system of record.
A Notable Prediction
Nick’s “unconventional” prediction is that software usage may actually decrease at the UI level.
What that means
- Customers may interact less with the app directly.
- Third-party and external agents may increasingly interface with software on behalf of users.
- Natural language and agentic workflows may replace many point-and-click interactions.
So software companies need to prepare for a world where:
- their users don’t log in as often,
- but their business logic is still being used heavily through agents.
Key Takeaways
- AI adoption in software development is no longer optional.
- Engineers should increasingly start in agentic tools, not traditional IDEs.
- Product, design, and engineering should collapse into tighter, earlier collaboration.
- Measure teams by customer and business outcomes, not just output.
- Put agents on the org chart and manage them like teammates.
- Build internal AI skills into portable “skills” files and guardrails.
- Rewrites are more feasible now, but only after defining the future platform.
- Product moat still matters, but it now includes education, data, and ecosystem value.
- Expect software usage patterns to shift toward agent-driven interactions.
Practical Recommendations for Founders
If you’re running a vertical SaaS company, Nick’s advice implies a clear action list:
- Audit your team: Who is actually using AI every day?
- Update expectations: Make AI use part of the job description.
- Create AI pods: Pair product, engineering, and AI operators on real work.
- Codify internal expertise: Turn your best reviewers, testers, and operators into agentic workflows.
- Stop measuring only velocity: Focus on outcomes and customer value.
- Plan for agentic interfaces: Design for the world beyond the traditional UI.
- Stay tool-agnostic where possible: Build portable workflows so you can adapt as vendors shift.
If you’d like, I can also turn this into a shorter executive summary or a bullet-only version for quick reading.