The Tech Trek

Mortgage is often the largest financial transaction in a person’s life, yet much of the process still runs on outdated workflows, long waits, and paperwork heavy systems.

Diane Yu, cofounder and CEO of TidalWave, explains how her team is rethinking mortgage technology with a twenty four seven AI loan assistant built for real work, not just demos. She shares why agentic software in financial services needs auditability, domain knowledge, and a different mindset around human oversight.

This episode is a sharp look at what AI adoption looks like inside mature industries where trust cannot be assumed and mistakes carry real consequences.

Key Takeaways

• AI in regulated industries cannot just be a generic model wrapper. It has to understand the domain, the rules, and the workflow deeply enough to operate safely.

• Diane makes a useful distinction between human in the loop and human on the side. The goal is not to remove people, but to let AI do the busywork while humans review, guide, and make judgment calls.

• Transparency is the foundation of trust. In mortgage, that means logging interactions, making AI recommendations visible, and giving loan officers enough context to validate the work.

• Productivity gains matter most when they free professionals from repetitive tasks and give them more time for borrower relationships, sales, and higher judgment work.

• AI native companies also have to use AI internally. Diane explains how TidalWave uses AI to move faster as a small engineering team and release product updates at a pace larger competitors may struggle to match.

Timestamped Highlights

00:33 Why mortgage is still stuck in outdated technology, and how TidalWave is building an AI loan assistant for the industry

01:55 Why incremental tools are not enough to fix a process that takes weeks, hundreds of pages, and still loses money per loan

03:46 The shift from workflow software to a system of action where AI does the work and prepares it for human review

05:53 Why regulated industries create a high bar for AI, including TidalWave’s mortgage compliance benchmark work with Columbia University

08:16 How trust is built through transparency, audit logs, human review, and clear boundaries on what AI should and should not answer

12:33 How AI can give mortgage professionals more time for human judgment, borrower relationships, and higher value work

22:12 How TidalWave monitors new AI models and uses AI internally to accelerate engineering speed

A Line That Sticks

“Human connection is what AI cannot replace.”

Pro Tips

• Do not treat AI as a feature button. Build it into the flow of work so it can actually remove friction.

• In regulated markets, design for auditability from the start. Every interaction, recommendation, and handoff needs to be visible.

• Use AI to improve throughput, but keep humans focused on judgment, trust, and relationship building.

• Watch model development closely, but build your platform so you can evaluate and plug in the best models over time.

Call to Action

Follow The Tech Trek for more conversations with the builders, operators, and technical leaders shaping how AI is changing real industries. If this episode gave you a useful way to think about agentic AI, share it with someone working through similar questions.

Accounting has a tech problem, and it is not what most people think. In this conversation, Cos Nicolaescu, Co-founder and CEO at Accrual, breaks down why accounting has lagged behind, where AI can actually create leverage, and why the future of the profession is likely more human, not less.

This episode gets into the real constraints inside accounting workflows, the difference between deterministic work and judgment based work, and why trust, context, and client knowledge still matter more than most AI narratives admit.

What stands out

• Accounting has not been slow to adopt tech because accountants resist change. A big reason is that the tools have often been weak, fragmented, and not worth the workflow overhead.

• AI can handle more of the mechanical and backward looking work, but the highest value still sits in judgment, context, and forward looking decisions.

• In accounting, knowing the tax code is not enough. The hard part is knowing the client, their history, their complexity, and the tradeoffs that shape future outcomes.

• The profession is still deeply supply constrained. Firms are understaffed, demand is growing, and better tooling may help accountants do more meaningful work instead of simply shrinking headcount.

• Junior talent may benefit more than people expect. As tools improve, newer professionals could ramp faster, though trust and client relationships will still take time to build.

Timestamped highlights

00:39 What Accrual is building, and why accounting workflows are a major opportunity for AI

01:47 Why accounting is more tech conservative than people assume, and why bad software is a big part of the story

04:11 The three buckets of accounting work, standardization, firm level process, and personal preference

07:29 Where AI fits best in accounting, flexible interfaces, document understanding, and smarter workflow support

12:14 The key difference between software engineering automation and accounting automation

16:36 Will AI reduce the need for accountants, or make the profession more productive and more valuable

A line worth remembering

“I would want to see people who spend a lot of time getting CPA degrees and training for decades spending most of their time, not just inputting data from one field to another.”

Pro tips

• If you are building AI for a regulated or detail heavy workflow, start with where accuracy matters most and do not confuse automation with value

• If you work in professional services, context is the moat. The more client history and situational knowledge you can capture, the stronger your systems become

• If you are early in your career, tool fluency can compress the learning curve, but trust still has to be earned

Stay connected

If this episode gave you a new lens on AI, accounting, and the future of expertise, follow the show, subscribe for more conversations like this, and share it with someone building at the intersection of software, operations, and professional services.

What does it take to modernize healthcare infrastructure when uptime is not just an SLA, but a patient outcome?

In this episode, Amir talks with Jeff Sponaugle, CTO of Surescripts, about building and operating mission critical healthcare systems, navigating the move from on premises infrastructure to the cloud, and figuring out where AI can create real value without compromising reliability. It is a sharp conversation on engineering judgment, modernization, workforce evolution, and why technical leadership still needs real technical depth.

What stood out

Cloud migration in healthcare is not just a cost or architecture decision. It is a reliability decision with real downstream impact on patients.

The best reliability strategy is not pretending nothing will ever break. It is designing systems so the customer never feels the break.

In regulated industries, structure can be an advantage. Standardized data and consistent formats make AI more useful, especially in healthcare.

AI can already improve the patient and clinician experience in practical ways, from transcription to summarizing complex records and surfacing relevant context faster.

Technical leaders cannot afford to drift too far from the work. Jeff makes the case that strong CTOs stay close enough to the technology to understand the tradeoffs, guide teams well, and spot what matters next.

Timestamped Highlights

00:00
Jeff Sponaugle joins the show to unpack mission critical technology in healthcare, cloud migration, AI, and workforce upskilling.

01:57
Why Surescripts sits in a critical layer of healthcare, and why reliability matters when prescriptions need to move in real time.

04:02
A simple but powerful view of reliability: things will break, but the customer should not know they broke.

06:47
How to adopt new technology without risky hard cutovers, and why parallel systems matter in high stakes environments.

08:53
Upskilling legacy teams, preserving tribal knowledge, and why continuous learning matters more than any single technical skill.

11:58
How regulation can actually help AI in healthcare by creating more consistency in the data.

17:33
Where AI and agentic systems could create meaningful value in prescribing, diagnostics, and clinical workflows.

20:29
Why AI has changed executive and boardroom conversations in a way cloud migration never did.

A line worth remembering

“The customer should not know that something broke.”

Pro Tips

If you are modernizing a high stakes platform, avoid the big overnight cutover. Run systems in parallel where possible and learn behind the scenes before customers ever feel the change.

If you lead technical teams, do not treat upskilling as a one time event. Give people a path to split time between legacy work and emerging systems so the transition is real and sustainable.

If you are evaluating AI in a regulated environment, start with narrow, useful workflows where context, speed, and summarization matter, then expand from there.

Stay connected

If you enjoyed this episode, follow the show, subscribe wherever you listen, and share it with someone building in healthcare, cloud infrastructure, or AI. You can also connect with Amir on LinkedIn for more conversations at the intersection of technology, leadership, and the future of work.

Farzan Karimi, Deputy CISO at Moderna, joins Amir Bormand for a sharp conversation on one of the most misunderstood areas in cybersecurity, the ethics of offensive security. From red team rules of engagement to nation state deception and the limits of AI in security testing, this episode gets into what happens when the job requires you to think like an attacker without crossing the line.

This is a practical conversation for security leaders, engineers, and operators who want a clearer view into how modern security programs actually work under pressure. Farzan shares hard lessons from his own career, explains why red teaming is really about business risk, and makes the case for storytelling over dashboards when security teams need executive buy in.

Key Takeaways

• Offensive security is not about finding every weakness. It is about simulating what a real attacker would do to reach the business’s worst case scenario.

• The gray area is real. Just because you are authorized to test a system does not mean every possible action is justified.

• Nation state level threats force teams to think differently. Attackers look across the connective tissue of systems, not just isolated tools or apps.

• Good red teaming can make the rest of the business stronger by helping teams see real risk, align on priorities, and justify investment.

• AI can speed up security work, but it still misses too much to replace experienced human operators.

Timestamped Highlights

02:02 What offensive security actually means, and why the best programs are built around business impact, not just technical findings.

03:46 Where the ethical gray area starts, from phishing and social engineering to the personal judgment calls that can end careers.

06:03 A story from Farzan’s Microsoft days that shows how a valid finding can still go too far when judgment slips.

11:06 Why security leaders have to explain to executives that attackers do not care about internal process, approvals, or red tape.

14:46 A nation state honeypot turned the red team into the target, and forced a complete shift in approach.

24:14 AI is changing the workflow, but Farzan explains why current tools still fall short of real red team depth.

A line worth remembering

“Just because you can doesn’t mean you should abuse those permissions.”

Pro Tips

• Tie offensive security work to the business’s real doomsday scenario, not a generic list of vulnerabilities.

• When you find a serious issue, know exactly where the rules of engagement stop, and stop there.

• Use attack stories and patterns to earn trust internally. Raw metrics rarely move people the same way.

• Treat AI as an accelerator, not a replacement for experienced security judgment.

Listen and follow

If this episode gave you a better lens on how modern security teams think, subscribe to The Tech Trek, follow the show, and share this episode with someone building, securing, or scaling technology in the real world.

Adi Kuruganti, Chief AI and Development Officer at Automation Anywhere, joins Amir to break down what it actually takes to build agents for the enterprise, not in theory, but in environments where complexity, governance, observability, and real business outcomes matter.

This conversation gets into the part of enterprise AI that most people skip. Not just what agents can do, but what changes when you have to deploy them across regulated systems, measure performance in production, manage model drift, and rethink how product and engineering teams ship software. It is a smart look at where enterprise AI is going, and what technical leaders need to understand before the market catches up.

What stood out

• Enterprise agents are only as strong as their data, context, and deployment model. In large companies, that means dealing with hybrid environments, air gapped systems, privacy controls, and process level context, not just model quality.

• AI is changing more than coding. Adi explains how his team is using AI across the full software development lifecycle, from spec creation and test generation to production event triage and release workflows.

• The release process is shifting from periodic launches to continuous iteration. That puts more pressure on observability, because teams now have to track model behavior, latency, and runtime performance as features roll out.

• Security can no longer sit off to the side. Prompt injection, shared tenant risk, and post production anomaly detection all require security teams to work much closer to AI and product teams.

• Mass adoption is not just a technology problem. The tools are improving fast, but enterprises still need change management, clear use cases, internal operating models, and people who know how to make AI part of daily work.

Timestamped Highlights

00:00 Adi Kuruganti joins the show to unpack what enterprise agent development really looks like today, from deployment models to governance to observability.

02:07 Why enterprise agents are different. Adi explains why context, data control, and environment complexity matter more in large organizations.

04:57 How AI is reshaping the software development lifecycle. From code suggestions to automated tests to incident triage, AI is moving deeper into product delivery.

10:13 The old handoff model is breaking. Product, design, and engineering are starting to work in a much more fluid, AI assisted way.

12:22 What changes in release management when AI writes part of the code and teams ship continuously instead of waiting for big release cycles.

18:17 How enterprises should judge agent performance, from human review and exception handling to evals, runtime benchmarks, and model drift.

27:21 Adi on the real AI adoption curve, job disruption, and why the bigger shift is not replacement, but making AI part of how people actually work every day.

A line worth sitting with

“AI should be a core element of how they work.”

Worth applying

• If you are building with AI, evaluate more than accuracy. Cost, latency, and consistency matter too.

• If you are leading teams, do not treat observability as a nice to have. Runtime visibility is part of the product now.

• If you are thinking about adoption, start with a real business problem and scale from early wins instead of trying to automate everything at once.

Follow the show for more conversations with the builders, operators, and technology leaders shaping how modern companies are actually being built.

What happens when software engineers stop thinking like coders and start thinking like orchestrators?

In this episode, Amir sits down with Scott Gale, CTO and Founder of Fluency, to unpack one of the biggest shifts happening in engineering right now: the move from writing code by hand to directing AI agents with context, judgment, and intent. Scott shares how his team is already using coding agents in production, what that means for hiring and team design, and why the engineers who adapt fastest will be the ones who gain leverage, not lose relevance.

This conversation gets into the real change beneath the AI hype. Not just better tools, but a different shape of engineering work. Less manual syntax, more planning, auditing, collaboration, and system level thinking.

Key Takeaways

• The value of an engineer is shifting away from typing code and toward directing intent clearly

• Teams that give AI better context can get dramatically better output from coding agents

• Engineers do not need to become people managers, but they do need to learn how to manage agent driven work

• Hiring is starting to favor people who can collaborate, learn the product, and work effectively with AI

• Faster software delivery does not mean less to build, it often means companies can finally tackle more of the backlog

Timestamped Highlights

00:01 Scott Gale, CTO and Founder of Fluency, joins Amir to break down the shift from builder to orchestrator in modern engineering

02:36 How Fluency introduced coding agents with a three part approach: safe experimentation, mindset shift, and stronger context

04:35 Is this just the next step in software engineering, or does AI fundamentally change the role?

08:16 Why some engineers resist AI tools, and what helps people move from skepticism to real adoption

11:26 How technical interviews are changing as AI becomes part of everyday engineering work

16:59 Scott on whether companies will actually need fewer engineers, and why the demand for meaningful work is not going away

21:09 The practical lesson teams miss: better structured systems and better context make coding agents far more effective

One line worth remembering

“It’s not about losing your craft. It’s about managing a workforce of junior agents.”

Practical edge

Scott shares a useful operating principle for teams already experimenting with AI in engineering: if you want better output, do not start with prompts alone. Start with structure. The more clearly a system is organized, and the more context an agent can access, the more useful and reliable the result becomes.

That applies to hiring too. Technical skill still matters, but the engineers who stand out now are the ones who can collaborate across product and engineering, understand the business context, and make good decisions with AI in the loop.

Call to Action

If you are thinking through what AI means for engineering careers, team design, or product velocity, follow the show and share this episode with someone building in this new environment. For more conversations with founders and operators shaping where tech is headed, connect with Amir on LinkedIn.

What happens when software engineers stop thinking like coders and start thinking like orchestrators?

In this episode, Amir sits down with Scott Gale, CTO and Founder of Fluency, to unpack one of the biggest shifts happening in engineering right now: the move from writing code by hand to directing AI agents with context, judgment, and intent. Scott shares how his team is already using coding agents in production, what that means for hiring and team design, and why the engineers who adapt fastest will be the ones who gain leverage, not lose relevance.

This conversation gets into the real change beneath the AI hype. Not just better tools, but a different shape of engineering work. Less manual syntax, more planning, auditing, collaboration, and system level thinking.

Key Takeaways

• The value of an engineer is shifting away from typing code and toward directing intent clearly

• Teams that give AI better context can get dramatically better output from coding agents

• Engineers do not need to become people managers, but they do need to learn how to manage agent driven work

• Hiring is starting to favor people who can collaborate, learn the product, and work effectively with AI

• Faster software delivery does not mean less to build, it often means companies can finally tackle more of the backlog

Timestamped Highlights

00:01 Scott Gale, CTO and Founder of Fluency, joins Amir to break down the shift from builder to orchestrator in modern engineering

02:36 How Fluency introduced coding agents with a three part approach: safe experimentation, mindset shift, and stronger context

04:35 Is this just the next step in software engineering, or does AI fundamentally change the role?

08:16 Why some engineers resist AI tools, and what helps people move from skepticism to real adoption

11:26 How technical interviews are changing as AI becomes part of everyday engineering work

16:59 Scott on whether companies will actually need fewer engineers, and why the demand for meaningful work is not going away

21:09 The practical lesson teams miss: better structured systems and better context make coding agents far more effective

One line worth remembering

“It’s not about losing your craft. It’s about managing a workforce of junior agents.”

Practical edge

Scott shares a useful operating principle for teams already experimenting with AI in engineering: if you want better output, do not start with prompts alone. Start with structure. The more clearly a system is organized, and the more context an agent can access, the more useful and reliable the result becomes.

That applies to hiring too. Technical skill still matters, but the engineers who stand out now are the ones who can collaborate across product and engineering, understand the business context, and make good decisions with AI in the loop.

Call to Action

If you are thinking through what AI means for engineering careers, team design, or product velocity, follow the show and share this episode with someone building in this new environment. For more conversations with founders and operators shaping where tech is headed, connect with Amir on LinkedIn.

Spencer Penn, Co founder and CEO of LightSource, joins The Tech Trek for a sharp conversation on AI native procurement, agentic workflows, and what actually happens to knowledge work as automation gets better. This episode is worth your time because it moves past lazy takes about AI replacing jobs and gets into something more useful, how work changes, where human value holds, and why procurement may be more strategic than most companies treat it.

This conversation starts with procurement, but it quickly expands into a bigger discussion about role design, change management, and the pace of AI adoption inside real companies. Spencer breaks down why some jobs get redesigned while others disappear, how AI can elevate overlooked functions, and what people should do right now if their company is behind.

In this episode

• Why procurement is a strong fit for AI, especially where teams are buried in tedious process work
• The difference between job automation and job elimination
• Spencer’s idea of role plasticity, and why it matters more than most AI debates
• Why procurement teams may become more valuable, not less, as AI improves
• Practical ways professionals can start using AI before their company rolls out a formal strategy

Timestamped highlights

• 00:37 What LightSource does and why direct material sourcing is a high stakes AI use case
• 01:51 Why procurement teams spend too much time on transactional work
• 06:47 Which jobs get enhanced by AI, which ones get eliminated, and Spencer’s framework for role plasticity
• 13:44 What the next few years could look like for procurement professionals
• 26:18 Where to start if your company has not adopted an AI native workflow yet
• 30:07 How to learn more about LightSource and connect with Spencer

“AI will not replace your job. Someone who knows how to use AI will.”

A practical thread running through this episode is simple. Start using the tools now. Use foundation models for secondary work, reporting, summaries, and internal communication. Build familiarity before the workflow shift gets forced on you.

If you are interested in AI, procurement, operations, supply chain, or the future of knowledge work, follow The Tech Trek for more conversations like this.

Michael Fanning, CISO at Splunk, joins The Tech Trek for a grounded conversation on how the security leader role is changing in the AI era. This episode gets into the real tension facing modern CISOs, balancing risk without slowing the business down, hiring for technical depth over narrow credentials, and defining success in a field where perfection is not a realistic metric.

This is a practical conversation for security leaders, engineering leaders, founders, and operators trying to make sense of AI adoption inside the enterprise. Mike breaks down why security has to move from fear based messaging to business enablement, why many teams may be overlooking strong security talent hiding in adjacent technical roles, and where AI can either reduce burnout or make it worse.

In this episode

• Why the CISO role is becoming more engineering driven and more tightly tied to business outcomes

• Where AI creates real leverage for security teams, and where it introduces new operational risk

• Why the security talent gap may be as much a hiring mindset problem as a supply problem

• What actually causes burnout in security teams, beyond the usual talking points

• How to think about success in security when zero incidents is not a serious metric

Highlights

• 1:44, The CISO role is shifting from pure protection to business enablement

• 7:11, AI creates leverage for defenders, but it is also accelerating the attacker playbook

• 9:31, The biggest AI security risks, from developer copilots to agent driven decision making

• 14:15, Why security teams need room to experiment with AI or risk falling behind

• 16:58, Only 1 percent of CISOs surveyed prioritized technology to close the skills gap

• 22:16, AI can reduce burnout, but only if it cuts noise instead of creating more of it

Security is about assessing risk and finding a way to say yes in a way that is responsible.

A practical idea worth taking back to your team
Look beyond candidates with formal security titles. Mike makes the case that strong engineers, SREs, and cloud practitioners often already understand the systems, access models, and infrastructure realities that matter most. Security can be taught on top of that foundation.

Link to report: https://www.splunk.com/en_us/form/ciso-report.html

Follow The Tech Trek for more conversations with leaders shaping how technology actually gets built, secured, and scaled.

What does it really take to go from engineer to CEO?

In this Tech Trek Brief, Michael White, Co founder and CEO of Multiply, shares a few of the ideas that matter most from a broader conversation on founder growth, leadership, and the shift from building things to building a company.

What stood out most is that this is not really a story about title progression. It is a story about learning to operate with more uncertainty, taking on bigger challenges before you feel ready, and realizing that leadership at the highest level starts to look a lot more like influence than execution.

What we get into

• Why growth often starts before you feel ready

• Why strong founders are pulled by a real problem

• Why founder timing matters more than people think

• Why leadership becomes influence, alignment, and conviction

Timestamped highlights

00:00 The real shift from engineer to CEO

00:18 Growth starts before readiness

00:56 Leadership changes when execution is no longer enough

01:50 The best founders are pulled by a problem

02:35 The three ideas that tie it all together

Follow The Tech Trek for more conversations on leadership, company building, and the people shaping what comes next.

The full Michael White episode is also available.