The Tech Trek

AI is not just changing software. It may also change the economics of the services businesses built around it.

Anirudh Sriram, CTO at Tessera Labs, joins The Tech Trek to explain how his company is using AI to take on enterprise transformation work traditionally handled by large systems integrators. Tessera focuses on migrations, ERP upgrades, code, data, and planning, but the bigger story is how a startup can compete by replacing large teams and long projects with automation, smaller teams, and a focus on outcomes.

The conversation also gets into a harder problem. AI can produce work much faster than people can verify it. In one example, Tessera completed code migration work in three days, but functional testing still required roughly two months. That gap between production and verification may become one of the biggest constraints on enterprise AI.

What Stood Out

• AI creates an opening for startups to compete in markets where incumbents have historically won through scale and headcount.
• Selling outcomes instead of large project teams can change both pricing and customer expectations.
• Enterprise migrations can be a wedge into a much larger opportunity because they require understanding a customer's systems, data, code, and business processes.
• Faster AI output does not remove the need for human review. In some cases, verification becomes the new bottleneck.

Key Moments

02:12 Where AI can take work out of the enterprise migration process
04:39 How transformation projects can stretch years beyond their original plan
08:27 Why AI may change the economics of services businesses
13:42 Why verifying AI output is becoming a major constraint
22:17 How Tessera approaches security, governance, and enterprise data
24:56 Using migration as the entry point into broader enterprise automation

One Line That Stuck

“We sell the outcome and not the process.”

Follow The Tech Trek for more conversations on building, operating, and competing with AI.

AI changes more than the product roadmap. It changes how engineering teams build, how data flows through the company, and what a CTO needs to own.

Andrew Rabinovich, CTO and Head of AI at Upwork, joins The Tech Trek to talk about his move from leading AI into the broader CTO role. His view is simple: AI is no longer just another component inside a software system. Increasingly, AI is the system, and infrastructure, data, engineering, and product need to be designed around that reality.

Andrew explains how that shift is changing Upwork's product development, from adding AI to individual features to building systems that learn across the entire user journey. He also discusses faster iteration, the importance of real time data, and how software engineering changes when machines can generate most of the code.

What Stood Out

• AI first development requires thinking about the entire system, not adding AI capabilities to isolated product features.
• Product iteration can move from months between versions to daily updates when systems continuously learn from user interactions.
• Engineers are moving from writing every line of code toward reviewing, steering, simplifying, and evaluating machine generated code.
• Asking the right question and knowing when a result is good enough may become more valuable than the mechanical work between those two points.

Key Moments

03:39 AI moves from being a component of software to becoming the foundation of the system.
07:35 How an AI background changes the role of the CTO and the relationship between technology and product.
10:48 Why Upwork moved from AI inside individual features toward end to end learning across the user journey.
16:04 AI makes feature creation easier, but faster creation does not automatically mean adoption.
20:41 Software engineering shifts toward reviewing and steering code generated by AI agents.
27:01 The skills that become more valuable when AI handles more of the execution.

One Line That Stuck

“AI is no longer a component of a large software system. AI is the system.”

Follow The Tech Trek for more conversations on AI, engineering, product, data, and technical leadership.

AI is making software faster and cheaper to build. It is not making trust any easier to earn.

Simon Wu, Partner at Cathay Innovation, joins The Tech Trek to discuss how AI is changing investment opportunities across healthcare, fintech, insurtech, legal services, and other regulated industries.

The conversation looks at a major shift in software economics. Instead of simply selling seats, licenses, and tools, AI companies can increasingly perform more of the work and deliver the outcome a customer actually wants. That creates opportunities for new business models, especially in industries where customization and services historically made it difficult to scale.

Regulation adds another dimension. Compliance, accuracy, governance, and complex workflows make these markets harder to enter. But Simon argues that the same friction can create defensibility once a company earns the trust of its customers.

Key Takeaways

• Regulation can become a moat. AI may lower the cost of building software, but companies still have to earn the right to operate inside sensitive workflows.
• Software is moving closer to outcomes. Customers increasingly care about the result, not how many seats or licenses they purchased.
• AI changes the economics of customization. Companies may no longer have to choose as sharply between scalable software and labor intensive services.
• Human involvement still matters. In healthcare, wealth management, and legal services, AI can make professionals more efficient without requiring them to disappear from the workflow.

One Line That Stuck

“AI has dramatically lowered the cost of building software. It doesn't lower the cost of earning trust.”

Follow The Tech Trek for more conversations on AI, engineering, product, data, and building modern technology companies.

AI can make individual tasks faster while leaving the organization with the same old coordination problems, or even making them worse.

Sergei Sorokin, CEO and co founder of Highlight, joins The Tech Trek to discuss why faster output does not automatically mean better work. Teams can generate documents, code, notes, and analysis faster, then spend the time they saved reshaping that output, moving information between tools, and figuring out what matters.

The bigger problem, Sergei argues, is often not access to information or model intelligence. It is context. AI needs to understand what matters to a specific person, team, and moment rather than simply searching across everything available.

The conversation also covers proactive AI assistants, privacy and security, team specific customization, and why trust will shape how quickly people allow AI to act on their behalf.

Key Takeaways

• Faster task completion does not eliminate the coordination tax between people and tools.
• The challenge is increasingly signal versus noise. AI needs to understand which information matters now.
• AI that adapts to individual teams could help companies preserve what makes their work distinct rather than producing increasingly similar output.
• Adoption will depend on trust. Drafts, approvals, undo options, and clear boundaries can help people become comfortable giving AI more control.

Key Moments

02:09 Why faster AI output can still create more work across teams
05:02 The coordination tax that existed before AI and why AI can amplify it
08:19 Why chat alone may not be the right starting point for workplace AI
13:58 How AI could adapt to teams rather than forcing teams to adapt to software
18:16 Why human behavior and trust will determine AI adoption
22:05 Why greater agent autonomy may create a demand for more user control

One Line That Stuck

“It’s not an intelligence gap. It’s a context gap.”

Follow The Tech Trek for more conversations about AI, engineering, product, data, and how technical teams are changing.

AI is changing hiring in ways that go far beyond candidates using ChatGPT to answer interview questions. The harder problem is knowing whether the person on screen is actually who they claim to be, whether their answers are their own, and what happens if someone with malicious intent gets access to company systems.

Yagub Rahimov, CEO and founder of Polygraf AI, joins The Tech Trek to discuss the growing trust problem surrounding AI assisted interviews, deepfakes, impersonation, and security. He explains why organizations need more visibility into the hiring process without turning every unusual behavior, accent, or response into a reason for suspicion.

What You’ll Take Away

• AI interview fraud is not simply a recruiting problem. Once someone enters the company, identity and access become security concerns.
• Detecting suspicious candidates based on human intuition alone can create false positives. Rahimov argues for using technology to create evidence and visibility.
• Small pieces of public information can reveal far more about a company than leaders realize when they are combined through what Rahimov calls mosaic intelligence.
• Protecting company information means thinking beyond traditional security controls to what employees, executives, and systems expose publicly.

Key Moments

02:27 How AI tools can turn legitimate technology into an interview cheating mechanism
05:35 Why hiring fraud can become a security and data access problem
07:43 Using voice, conversation context, and AI detection to improve visibility during interviews
11:43 Why increased AI uncertainty should not lead companies to distrust everyone
14:28 What organizations should think about after a candidate actually gets hired
19:40 The continuing race between increasingly capable deepfakes and detection technology

One Line That Stuck

“Tech problems have tech solutions.”

Follow The Tech Trek for more conversations about AI, engineering, data, product, and how technical teams are adapting.