Product Thinking – Episode 241: Mastering AI Strategy in Enterprise Teams with Maryam Ashoori

Date: September 3, 2025
Host: Melissa Perri
Guest: Dr. Maryam Ashoori, VP of Product and Engineering, IBM WatsonX

Episode Overview

This episode explores how enterprise teams can master AI strategy in product development, guided by Dr. Maryam Ashoori’s deep experience at IBM WatsonX. The discussion moves beyond the AI hype, focusing on the importance of intentionality, responsible innovation, and technology-agnostic thinking. Maryam and Melissa break down AI agents in enterprise use, best practices for product leaders, common pitfalls, and the future convergence of product, engineering, and design roles.

Key Discussion Points & Insights

1. Defining AI Agents and Their Role in Enterprise (05:07–06:23)

• What is an AI agent?

  • An AI agent is "an intelligent system with reasoning and planning capabilities that can automatically make decisions and take actions.” (Maryam, 05:08)
  • Distinction from LLMs: AI agents use LLMs for reasoning/planning, but expand to tool and action-calling (e.g., summarizing podcasts, executing follow-up actions).

• Use Case Example:

  • Customer support for camera troubleshooting through retrieval-augmented generation – agent searches documentation, escalates to web search, then consults a human if necessary. (08:58–10:31)

• AI agents vs. LLM chatbots:

  • “There is some kind of discretion level programmed into these AIs.” (Melissa, 10:31)
  • Agents orchestrate multiple steps/tools, unlike single-turn QA models.

2. Building AI Products: Intentional Problem-Solving (02:43–04:48, 31:40–33:17)

• Maryam’s approach: Start with the business problem, not the technology.
  • “I go to the business problems and I question, is this problem well suited to be tackled by AI? And if so, what are the advantages, what are the risks?” (Maryam, 03:49)
• Caution against chasing tools: “Don’t be distracted by the technologies coming out every day... focus on the problem that you’re trying to solve and be very intentional about where you spend your time.” (Maryam, 31:52 & 01:10)

3. Risks, Limitations, and Guardrails of AI Agents (13:27–18:45)

• Productivity Gains vs. New Risks:
  • Up to 1–2 hours (sometimes 6) saved per developer daily using AI-assisted coding (Maryam, 13:27)
• Unique risks: AI agents take action—beyond generating content, they can “leak data” or even “wipe out your code.” (15:04)
• Guardrails: Build agentic guardrails (e.g., faithfulness to context, content filters) and mandate human oversight for sensitive information.
  • “As a product manager, you can put together these guardrails in the flow of decision making for the agent to make sure that they stay close to that truth.” (Maryam, 17:25)
• Calculated Risk: Weigh severity: “The higher the risk, the more you want to put these [guardrails] into place.” (Melissa, 18:45)

4. Understanding Hallucinations and Their Management (16:07–17:25)

• Why LLMs hallucinate:
  • “There is no reasoning, really, there is no logic of thinking behind LLMs. This is an unsupervised learning ... it basically calculates what's the probability of the next token.” (Maryam, 16:15)
  • Solutions: Guardrails, prompt engineering, and always having a human in the loop for high-stakes outputs.

5. Evolving Roles in Product, Engineering, and Design (22:10–26:23)

• Blurring boundaries:
  • Historic PM/engineer ratio (1:6–10) shifting due to AI; e.g., 1:0.5 cited (Maryam, 22:15)
  • “I’m less worried about the labeling of people ... the lines between these two roles are emerging big time with AI.” (Maryam, 22:36)
• Designers’ future: Focused on prompt engineering and UX strategy, not pixel-perfect UIs.
  • “Designers need to focus their expertise on asking the right questions and testing the right thing versus building the UI itself ... be the master of defining the right promptings for AI.” (Maryam, 23:23)
• The Risk of Uniformity:
  • With models trained on the same data, creativity and unique “designer touch” will set products apart. (Maryam, 24:53)

6. Strategic Adoption of AI in the Enterprise (29:03–34:38)

• Start with Small, Validated Use Cases:
  • “Start small, identify what is the problem that you’re trying to solve. I see a lot of POCs, proof of concept happening in the market.” (Maryam, 29:03)
• Scalability:
  • Consider performance, latency, energy requirements before scaling.
• Technology Agnosticism:
  • “Any decision that you make, try to be technology agnostic ... abstract away the business case from the technology part.” (Maryam, 33:17)
• Tool Chaos:
  • Developers report needing to stitch together 5–15 tools for AI applications (Maryam, 30:16)
• Coping with Rapid Change:
  • Don’t chase every new startup/tool; “You can’t get that time back.” (Maryam, 31:52)

7. Organizational Impact & Future Outlook (26:23–28:10, 34:44–36:35)

• AI as a Tool (Calculator Analogy):
  • “The skills that they are teaching students these days is how to leverage the calculator. Even the advanced calculations of calculators or maybe AI today to go solve bigger problems.” (Maryam, 26:23)
  • AI accelerates iteration, experimentation, and user testing, enabling “a lot more domain specific targeted applications... the cost of development is going low.” (Maryam, 27:09)
• Emerging Trends:
  • Four areas for product leaders:
    • Boosting productivity (community learning, sample prompts)
    • Enriching existing products (feature acceleration)
    • New business creation (LLM-powered startups)
    • OEM/embedded provider opportunities
  • “Depending on where you are operating, the path to take forward... is going to be very different.” (Maryam, 36:11)

Notable Quotes & Memorable Moments

• On adopting AI tech responsibly:
  “Don’t chase the solutions, take one step back and identify what is the problem that you’re trying to solve and be very intentional about where you spend your time.”
  – Maryam Ashoori (31:52)

• On AI hallucinations:
  “There is no reasoning, really, there is no logic of thinking behind LLMs ... it basically calculates what's the probability of the next token.”
  – Maryam Ashoori (16:15)

• On future roles:
  “I’m less worried about the labeling of people ... the lines between these two roles are emerging big time with AI.”
  – Maryam Ashoori (22:36)

• On uniqueness in design:
  “If you don’t tune [the models] or apply your nice designer touch... it’s going to generate the same design that someone else has generated. What is going to win... is the ones that are grabbing that and add their differentiated touch.”
  – Maryam Ashoori (24:53)

• Advice for her younger self:
  “Be very intentional about where to spend your time because the board changes but you don’t get that time back.”
  – Maryam Ashoori (36:49)

Timestamps for Important Segments

• 02:43 – Maryam’s career journey and AI philosophy
• 05:07 – What is an AI agent? Core definitions and capabilities
• 08:58 – Real-world enterprise example: AI agents in customer service
• 13:27 – Survey findings: real impacts, risks, limitations for engineers
• 16:07 – Why large language models hallucinate
• 17:25 – Building guardrails and mitigating risk in AI workflows
• 22:10 – Blurring lines: PM & engineering roles in the age of AI
• 23:22 – Evolving designer roles, focus on strategy and prompting
• 24:53 – The risk of UX homogeneity and the need for creativity
• 29:03 – Enterprise adoption: start small, scale wisely, stay tech-agnostic
• 30:16 – Tool fragmentation: the challenge of integrating multiple AI tools
• 31:40 – Strategies for tech adoption (focus, intentionality)
• 33:17 – Core components of an effective AI product strategy
• 34:44 – Key trends for product leaders in AI
• 36:49 – Advice to her younger self and today’s product professionals

Takeaways for Product Leaders

• Always start with the problem, not the technology.
• Use AI as a tool for acceleration, not as a replacement for thoughtful design, engineering, or product strategy.
• Build robust guardrails and include humans in the loop for critical decisions.
• Prepare for role convergence—engineers, PMs, and designers will collaborate more closely, with deeper overlaps.
• Stay tech-agnostic and intentional with your time and resources.
• Remain creative; don’t let models drive out distinctiveness in your products.
• Continuously learn from the community, monitor trends, and focus on business value.

To connect with Dr. Maryam Ashoori:
Find her on LinkedIn, especially regarding WatsonX Governance. (37:04)

For more:
Visit ProductThinkingPodcast.com for links and show notes.

This summary captures the substance, actionable guidance, and inspiration from the episode, enabling leaders and practitioners to chart a grounded and forward-thinking path for integrating AI into their teams and products.