Tech Talks Daily

What happens when marketing software starts learning from customer behavior instead of following a campaign rule that someone wrote weeks ago?

In this episode of Tech Talks Daily, I speak with Neha Mittal, co-founder and CEO of JustAI, about AI agents, reinforcement learning, and a different approach to personalization. Neha spent over a decade working on growth and retention at companies including Twitter and Pinterest. Her experience repeatedly brought her back to the same problem: teams were running sequential experiments while the interests and behavior of their users kept changing.

Neha describes how a series of small A/B tests could consume weeks without producing a durable answer. Changing a button from red to blue might teach a team something, but the larger opportunity was to keep experimenting and learning across many decisions. Seeing similar problems at different companies helped shape the idea behind JustAI.

We discuss why she considers marketing a promising environment for autonomous agents. It involves a high volume of decisions, frequent feedback, and experiments that can often be assessed quickly.

Opens, purchases, and ignored messages can all contribute to learning. Neha contrasts those conditions with higher-consequence fields such as medicine and law, while explaining that strategy and judgment remain responsibilities for people.

The conversation also separates generating content from choosing an action. A language model can help produce an email. The reinforcement learning approach Neha describes aims to learn which decision supports longer-term customer value. The question becomes what message, timing, or channel makes sense for a particular person in the current moment, rather than which static segment or campaign they belong to.

JustAI divides the work across Strategy, Creative, Decisioning, and Data agents. Neha explains how opportunities feed into messaging, how the decisioning system learns from responses, and how the data agent returns insights to the process. Marketers can review customer-facing creative, provide brand guidance, and adjust strategy when the company's products or priorities change.

Her hypermarketer vision raises another question about how teams grow. Neha imagines one marketer having the capacity traditionally associated with a much larger team, with agents handling repetitive execution and experimentation. She presents this as a direction for the business rather than a demonstrated productivity ratio for every customer. The marketer still needs to translate business goals into instructions and judge whether the results make sense.

We bring the discussion back to current examples. Neha describes Coursera and ClickUp delegating the creation and selection of messaging variants. She also discusses the importance of human review for Better's financial-services communications. These examples show why the level of autonomy should depend on the task and the consequences of getting it wrong.

The supplied booking notes report over $100 million in customer revenue influenced and a $17 million funding round. Those company claims provide background, but the interview does not establish a revenue attribution method, incremental lift, or a complete return-on-investment calculation. The practical value of the conversation lies in understanding what work is being delegated and what people continue to own.

For me, the interesting issue is how to create a feedback loop that improves relevance without letting operational speed run ahead of customer trust or business judgment.

Which marketing decisions would you delegate to AI agents today, and what would you want to review before a message reaches a customer? Please listen and share your thoughts. 5. Suggested URL

Is your company using AI to improve the buying experience, or to carry out the same processes faster?

In this episode of Tech Talks Daily, I speak with Jonathan Kvarfordt, known as Coach K, about business strategy, go-to-market AI, and the relationship between people and automated systems. Jonathan introduces himself as VP of Marketing at 1mind and brings experience training revenue leaders and teams to assess where AI can support their work.

We begin with his framework for optimizing, amplifying, or reinventing a process. The point is to recognize the difference between making an existing task easier and considering whether the task should be done differently. Jonathan also acknowledges that organizations have different levels of readiness. Changing everything at once is not a practical answer for every business.

The conversation becomes especially useful when we discuss measurement. Rather than beginning with licenses or prompt activity, Jonathan recommends identifying the business outcome and working backward. Is the goal improved quota attainment, a higher win rate, reduced churn, or another result? What is preventing it today? Sometimes the answer concerns process or people rather than AI.

We then turn to the customer journey. Jonathan describes how 1mind aims to help buyers get product information, receive demonstrations, and answer technical questions before moving into a conversation with a person. His argument is that a website visitor may already be researching a specific decision, so an experience built only around awareness and meeting bookings can leave their immediate questions unanswered.

The practical question is how to support that buyer at the moment they need help. Jonathan talks about website assistance, sales engineering, product guidance, and customer success. He sees opportunities to provide information when a team cannot staff every interaction, while giving people additional context when they take over.

We also discuss the human side of AI adoption. Jonathan describes the continued value of emotional intelligence, relationships, and strategy. He challenges the assumption that every buyer always wants to speak with a person, but also recognizes that providing facts is different from understanding someone's motivations. The opportunity depends on the individual, the task, and the trust they place in the experience.

A comparison between two companies using the same outbound tool brings the discussion back to operational reality. In Jonathan's account, one company had clearer customer definitions, messaging, and data. The other used a broad target and expected the technology to compensate. His lesson is to examine what the system receives before concluding that another purchase will solve the problem.

For leaders assessing AI value, we talk about conversion rates, selling time, and the parts of a buyer's journey a system can support. Establishing a baseline matters. A team cannot confidently claim improved selling time if it never measured how much time people had to sell in the first place.

The interview also contains an inconsistent percentage claim about hallucinations. I have kept those figures out of the promotional package pending clarification. Reliability needs evidence, and the discussion does not provide a benchmark methodology or enough detail to establish a performance guarantee.

We close with Jonathan's advice to remain willing to learn and reconsider familiar habits. He is candid about feeling overwhelmed by the pace of change, which makes the conversation useful for people who are experimenting without feeling they have every answer.

Where does your buying journey create unnecessary effort, and how would you assess whether AI made the experience better for the customer? Please listen and share your thoughts.

What would change if your building could help explain what needs attention, why it matters, and whether the action you took actually worked?

In this episode of Tech Talks Daily, I welcome Bert Van Hoof, CEO of Willow, back to the show to discuss operational AI for buildings and infrastructure. Bert spent about 20 years at Microsoft, where his work included launching Azure Digital Twins. He now works with environments such as hospitals, airports, universities, corporate campuses, and data centers, where the condition of a physical space can directly affect the work happening inside it.

We begin with a problem familiar to many facilities teams. Buildings already have sensors, management systems, meters, maintenance records, and dashboards. Having access to information does not necessarily help an operator decide what to do next. Bert describes operational AI as a way to move through a sequence of questions: what is happening, why is it happening, what might happen next, what should we do, and did the action produce the expected result?

Context gives those questions meaning. A temperature reading alone cannot tell us whether a room is operating correctly. We need to know the room's purpose, occupancy, equipment, weather conditions, and recent maintenance history. Bert explains how digital twins and knowledge graphs connect those relationships so that a system can assess the situation rather than interpret a sensor value in isolation.

The discussion then turns to predictive maintenance. A probability of failure may be useful to an analyst, but a technician also needs to understand the consequence, urgency, and next step. Bert describes how condition-based maintenance can support equipment decisions and how earlier detection can reduce the disruption caused by refrigeration problems or water leaks.

These are practical examples of the types of outcomes Willow targets, rather than a claim that every customer will achieve the same result.

There is a human side to the conversation too. Bert describes operators opening numerous browser tabs each morning to assemble a picture of their buildings. Daily briefings and natural-language questions could reduce that search effort, helping teams focus their attention on the issues that deserve it.

We also discuss the relationship between occupancy, ventilation, energy consumption, comfort, and maintenance. Bert describes Willow's Active Control and the use of schedules and occupancy signals to help buildings respond to how spaces are being used. His argument is that these decisions should account for the whole environment. Saving energy in isolation can create different problems if comfort, reliability, or safety is overlooked.

Trust runs through the interview. A poor recommendation in a hospital has different consequences from an awkward paragraph in a document. Bert argues for explainable recommendations, operational evidence, defined limits, and verification after an action. He also sees experienced facilities engineers as part of the solution, with responsibility for setting objectives and judging trade-offs.

For me, the larger question is how building operations connect with the organization's purpose. A hospital needs spaces and equipment ready for patient care. An airport needs to account for passengers, flights, gates, and changing demand. Understanding the environment means understanding what it needs to support.

Where would operational AI make the most useful difference in your buildings, and what evidence would you want before allowing it to take action? Please listen and share your thoughts.

How much damage can a proven leadership playbook cause when it is applied to the wrong company? Experience can give executives confidence, but Sameer Kazi argues that confidence becomes dangerous when it replaces diagnosis.

In this episode of Tech Talks Daily, I welcome Sameer Kazi, CEO of the Pantheon Platform, back to the show for a conversation about copy-and-paste management, first-principles thinking, AI adoption, and the value of listening before making changes. Pantheon supports major websites and provides DevOps tools for agencies and enterprise web teams, placing Sameer close to the changing relationship between SaaS, web development, and agentic website creation.

Sameer describes a familiar mistake in technology companies. A new executive arrives with a formula that worked elsewhere, hires people from a previous team, and assumes the same inputs will produce the same growth. In the early SaaS era, for example, a company might have treated its sales efficiency ratio as a direct signal to hire another quota-bearing sales representative.

Product-led growth, changing buying behavior, and different forms of product market fit have weakened that assumption. Adding salespeople does not automatically create demand.

His alternative begins with product market fit and the way buyers discover, evaluate, and purchase a product. From there, leaders can study the mechanics across marketing, sales, customer success, support, and product development.

Sameer argues that rapidly growing AI businesses do not come with a settled operating manual, so executives need curiosity and evidence instead of a formula copied from the past.

Customer feedback provides one source of that evidence. Pantheon runs a public Slack channel where customers and partners speak directly with the company. Sameer says those conversations can reveal customer experience problems and, in some cases, signal a platform slowdown before observability tools detect it.

A small piece of feedback may not represent every customer, but it can give an experienced leader a reason to investigate immediately.

We also discuss AI return on investment and the growing cost of adding token consumption to an organization's existing workforce and infrastructure. Sameer does not claim to have one universal metric. Instead, he recommends looking at customer reactions to AI across sales, support, and internal processes. If AI makes a process faster for the company but worse for the buyer, customer evidence should reveal the problem.

The conversation then turns to data ownership. Sameer compares the hesitation companies once felt about moving data into the cloud with the speed at which some now provide business data to frontier AI labs. He expects small language models and open-weight models to give companies additional choices, including the option to run systems on their own hardware and isolate proprietary information where appropriate.

One of the strongest leadership examples comes from Sameer's arrival at Pantheon. He promised to speak with every employee and spent four months holding small roundtables. The format began with introductions and favorite movies before moving to a direct invitation: tell me what works, what needs attention, and what should be addressed first.

Sameer describes the result as a treasure trove of information from people closest to customers, prospects, products, and daily operations.

That approach also challenged the lazy assumption that long-serving employees represent outdated thinking. Sameer found experienced practitioners who had built agencies, developed large fleets of websites, and worked directly with customers.

Recognizing that knowledge can strengthen commitment, while dismissing it can leave employees disconnected and encourage quiet quitting.

Can leaders keep the principles that made them successful while treating every familiar method as a hypothesis to be tested? Listen to the episode and share how your organization balances executive experience with customer evidence and employee knowledge.

What matters most when you land in another country and your phone has no data? The cheapest plan may have looked like the obvious choice before departure, but priorities change quickly when you need a map, a translator, a ride to the hotel, or a way to tell your family you arrived safely.

In this episode of Tech Talks Daily, I speak with Alfonso Mata Villa, Head of Strategy at Holafly, about the growth of travel eSIMs and why they have moved from a specialist product to a far easier alternative to physical SIM cards and traditional roaming.

Alfonso explains that modern smartphones now make it possible to keep a normal number active while adding a separate data connection for an international trip. The result is a product that feels less like replacing a phone service and more like adding connectivity for a destination.

We talk through the practical experience for someone using an eSIM for the first time, including installation, activation, turning off the home data connection, and avoiding an unexpected roaming bill. Alfonso also describes Holafly's Always On benefit, which the company says gives customers one gigabyte of data each month so they can access basic services after arrival before buying another plan.

The apparent simplicity hides a demanding operating model. Holafly works across over 200 markets, where network quality, device behavior, carrier restrictions, hardware differences, and local regulation can all vary.

Alfonso explains why providers need strong operator relationships, redundancy, and the ability to move traffic when coverage becomes unreliable. His argument is that travelers should never have to see the complexity required to make a connection work.

That raises a useful question for anyone comparing travel eSIM providers. Price and data allowance matter, but what happens after the purchase? We discuss network reliability, data routing, latency, access to local services, privacy, and the speed of customer support when an installation or connection fails. Alfonso says Holafly treats multilingual support, unlimited data, transparent pricing, and reliable connectivity as parts of the product rather than extras.

The conversation also looks at the changing relationship between eSIM companies and mobile operators. Travel eSIMs compete with traditional roaming, but they still depend on operator networks. Alfonso sees an increasingly connected model in which operators supply infrastructure while global providers package connectivity for travelers and distribute it through airlines, travel agencies, hotels, online travel platforms, and financial services.

There is also a risk of leaving some travelers behind. Older phones may lack eSIM support, while carrier-locked devices and regional differences can create additional barriers. Alfonso argues that progress should be measured by whether the technology becomes simple enough for almost any traveler to use, rather than by technical sophistication alone.

If mobile connectivity can become another digital part of planning a trip, will travel eSIMs finally retire the anxiety and surprise costs associated with international roaming? Listen to the episode and share your experiences with eSIMs, roaming charges, and staying connected abroad.

Could your board explain where the organization's most valuable data resides, who can control it, and how quickly it could be recovered

In this episode of Tech Talks Daily, I speak with Adam Gale, CTO for Cybersecurity and Artificial Intelligence at NetApp, about why digital sovereignty is moving from compliance teams into boardroom discussions.

Adam argues that data has measurable business value, so decisions about its location, ownership, protection, and recoverability should be treated with the same seriousness as other financial risks.

We discuss how the US CLOUD Act, Europe's sovereignty agenda, and wider geopolitical tensions are changing cloud decisions. Adam explains that sovereignty extends beyond physical data location.

Hardware supply chains, software development, labor, access, operational control, and recovery all influence the level of sovereignty an organization may require. His advice is to treat sovereignty as a scale rather than impose the same controls on every workload.

The conversation also turns to sovereign AI. Adam describes why businesses and governments want greater control over training data, intellectual property, model activity, and audit records. He explains how poisoned data can produce plausible but unsafe outputs, and why controlled AI environments, traceable logs, and immutable records matter when systems begin making consequential decisions.

We then connect data visibility with business value. Adam shares how uncertainty over data location can affect a company acquisition and explains why insurers increasingly ask organizations to prove how information is protected and recovered.

He also describes the growing role of behavior analytics, automated responses, immutable storage, and storage-level ransomware protection in limiting damage and restoring trusted data.

If digital sovereignty begins with responsibility, control, and security, who owns those decisions in your organization today?

Listen to the conversation and share your thoughts with me.

Why do promising AI pilots struggle when they become services that real customers depend on?

In this episode of Tech Talks Daily, I'm joined by David Hughes, Chief Technology Officer at Stelia, to talk about the engineering decisions behind enterprise AI deployment. Dave's responsibilities cover networking, cloud infrastructure, and customer-facing products, giving him a view of what happens when an application meets the physical systems expected to support it.

We begin with a question that can get lost in the excitement around a successful demonstration. What problem was the system designed to solve, and what happens when the number of users changes? Dave describes reviewing a proof of concept that had worked for one external customer, only to identify a messaging component he believed would become a bottleneck under heavier demand. His lesson is to understand the intended workload before treating a working prototype as a production service.

The discussion also offers a balanced view of cloud computing. Dave praises elastic resources and programmatic control, while arguing that demanding AI workloads require teams to understand how storage, networking, and compute interact. A fast network cannot compensate for storage that cannot feed it. His steamship analogy makes the point familiar: the engine needs a steady supply of fuel if it is going to maintain its speed.

We talk about lessons from high performance computing, Dave's experience at Genesis Cloud, and the practical constraints of finding data center space and power. He explains why location and available capacity can influence architecture before a team even starts comparing GPUs.

For business leaders, the conversation becomes especially useful when we turn to measurement. Dave describes assessing internal model services through their effect on spending and judging customer projects by their output. His hotel occupancy analogy raises a question for anyone approving an infrastructure budget: how much of the asset's capacity will actually be used?

There is also a tension between engineering discipline and experimentation. Dave describes starting an internal inference service with existing servers and GPUs, then expanding it as demand grew. Proving that people wanted the service helped establish a case for a fuller investment. The lesson is to distinguish a useful experiment from evidence that a system is ready for wider demand.

We close with the Stelia newsroom and its engineer-written technical content. The supplied guest notes also introduce Stelia AI OS, described by the company as a full-stack operating system for enterprise AI. The interview focuses on deployment principles rather than a detailed assessment of that platform.

Where does your organization encounter the biggest obstacle when moving AI into production, and what evidence helps you decide where to invest next? Please share your thoughts.

Can a hotel modernize its technology without turning hospitality into another automated transaction? In this episode of Tech Talks Daily, I speak with Mark Hollyhead, Chief Transformation Officer at Aven Hospitality, about the systems behind thousands of hotels and the operational decisions that affect both employees and guests. Mark brings experience from British Airways, Heathrow Airport, Expedia, and travel management before joining the company carved out from Sabre with support from TPG.

His argument begins with a distinction that many technology programs miss. Guests rarely remember the property management system, API, or data pipeline behind a stay. They remember whether somebody knew their name, solved a problem, explained when the room would be ready, or prevented them from repeating information after a long journey. Technology matters because it can remove the administrative work surrounding those moments. Mark wants frontline employees to have time to be hospitable, especially when a situation requires empathy or judgment. That means looking beyond the promise of convenience and asking whether a system gives staff better information, reduces repeated work, and leaves them free to focus on the person in front of them.

Mark shares an example from a hotel he visits regularly in Dallas. Different employees recognized a few preferences from previous stays, including his connection with Seattle, a request for two bottles of water, and a preference for a corner room. None of those details required the hotel to know everything about him. They gave the employee enough context to make the interaction feel personal.

The example also illustrates the boundary between thoughtful personalization and an experience that feels intrusive. Mark's test is whether the customer would be surprised by what the company knows. If the guest begins wondering how the information was obtained, trust has already weakened. Hotels hold identity, payment, preference, and location data, including the place where somebody is sleeping, so that responsibility deserves careful treatment. For technology investment, Mark begins with friction rather than a product category. He looks for employees rekeying information, building spreadsheets because systems do not communicate, switching among screens, and developing improvised techniques to complete ordinary work. Across one property, five unnecessary minutes may appear minor. Across thousands of employees and repeated tasks, those minutes become expensive.

This approach changes how transformation is measured. Mark says another login, dashboard, or screen usually indicates that software has been added without changing the underlying work. He asks what has disappeared after a release. Fewer clicks, calls, handoffs, and repeated explanations provide clearer evidence than the launch itself.

Hotels also depend on a large network of systems covering reservations, property management, payments, distribution, loyalty, revenue management, customer relationships, and housekeeping. When definitions differ across those systems, employees become the integration layer. Data quality problems then appear at the exact moment a guest expects a simple answer. Mark believes the arrival of agentic commerce will make trusted and consistent information even more important. Travelers who have spent decades filling out search forms may increasingly describe what they need through a conversation. The AI agent will expect accurate room attributes, availability, rates, preferences, and policies from connected sources. Conflicting answers will damage confidence quickly.

The episode also challenges the familiar debate between digital and human service. Mark argues that people do not wake up choosing one or the other. They want an easy experience suited to the circumstances. A traveler arriving after a normal flight may prefer to go directly to the room. A traveler stranded in a snowstorm is unlikely to appreciate a chatbot repeating that it understands their frustration.

His operating principle is to automate predictable work and make it simple. When a situation becomes ambiguous, emotional, or unusual, a person should be able to step in with the context already available. The handoff matters as much as the automation because the employee should not force the guest to repeat the entire story. This has implications for frontline adoption. Mark rejects the assumption that resistance means employees dislike change. People often dislike bad change, especially when a new system makes the job harder and the response is a lengthy training session explaining why it should feel better. Aven involves frontline staff earlier so their knowledge can influence workflows before technology is introduced.

AI can also help prioritize work by showing employees the next action that deserves attention. That does not eliminate human judgment. It can give staff the information and authority required to resolve the most important problem without searching through several systems while an increasingly frustrated guest waits.

The opportunity is a hospitality operation where technology works quietly, reliable information travels with the guest, and employees have greater capacity to create memorable moments. The risk is adding automation to disconnected processes, collecting personal information without a clear purpose, or mistaking another launch for an improved experience.

Where would removing five minutes of friction create the greatest difference for your employees and customers? Listen to the episode and share your thoughts.

When should an AI system be allowed to fix an IT problem without waiting for a human to approve the action?

In this episode of Tech Talks Daily, I speak with Matt Tuson, General Manager for Europe, the Middle East, and Africa at LogicMonitor, about the practical path from reactive IT operations toward autonomous IT. The promise is attractive: fewer repetitive tasks, faster incident resolution, less disruption, and systems that can correct familiar problems before users feel the impact. The difficult part is deciding when the available data, controls, and evidence are strong enough to trust an automated action.

Matt argues that observability is taking on a different responsibility. It once helped engineers answer what happened and why. In an autonomous environment, it must also give AI enough connected context to recommend or execute what should happen next. That means understanding how infrastructure, networks, cloud services, applications, and other operational signals relate to one another. A fast decision based on an incomplete view can resolve the wrong symptom, create duplicate incidents, or make the original problem worse.

Trust therefore begins with the information supplied to the system. Many IT environments contain separate monitoring and management tools introduced by different teams to solve specific problems. Those specialist products may continue to do useful work, but their signals often remain disconnected. Matt describes a business combining nine organizations, each bringing its own tools and operational practices. The challenge was not simply the number of products. It was the inability to connect their outputs into a reliable view of what was happening across the combined environment.

This matters because AI can process poor information as quickly as good information. Matt warns that adding autonomous remediation to incomplete, duplicated, outdated, or poorly contextualized data risks creating automated confusion. False positives multiply, root causes are missed, and teams repeatedly address symptoms while the underlying fault remains. His advice is to start with the business outcome, identify the information needed to support that outcome, and improve the quality and consistency of the data before increasing AI authority.

We also discuss how leaders can decide which actions belong with AI and which should remain under human control. Matt rejects the idea that organizations must choose between full autonomy and complete human approval. He recommends graduated levels of authority based on the action, confidence in the recommendation, business impact, and whether the change can be reversed easily. Ticket enrichment, alert correlation, and routine remediation may offer lower-risk starting points. A change affecting a major service should receive closer human oversight until the system has established a reliable record.

That creates a practical model for earning trust. If an AI recommendation repeatedly matches the action an experienced engineer would have taken, the organization gains evidence that it may be safe to automate that decision. The process can move gradually through support levels and operational complexity, with explainability, governance, and audit records maintained throughout. Human judgment remains where accountability and business consequences demand it.

Matt also addresses the difference between companies making progress with AI and those stuck in pilot purgatory. The stronger performers begin with an operational problem, such as reducing incidents, improving uptime, or accelerating resolution. They invest early in data quality, visibility, collaboration, and governance. AI becomes part of an existing workflow with a defined path from insight to action, rather than a detached experiment looking for a reason to exist.

For IT leaders who want greater autonomy but do not yet trust their environment, Matt's starting point is simple: establish broad visibility, improve operational data, connect isolated sources, and define the actions AI can take independently. Teams must also be able to understand why a system reached its conclusion and inspect what it did afterward. Autonomy should grow through repeated evidence, not through hope or one large leap.

There is a commercial reason to get this right. Faster resolution can reduce the cost of service, protect customer experience, and limit the business impact of downtime. There is also a human benefit. Engineers spend less time sorting duplicate alerts or joining war rooms to prove which team was innocent. They can concentrate on decisions that require experience, judgment, and knowledge of the business.

As observability starts supplying evidence to machines as well as people, what controls would make you comfortable allowing AI to take action inside your IT environment? Listen to the episode and share your thoughts.

What has to change before an AI pilot can become part of a regulated financial workflow?

In this episode, I speak with Tom Carey, President of Broadridge's Global Technology and Operations Business, about the work required to move AI beyond personal productivity tools and prototypes. Tom brings an unusual perspective to the conversation. He studied AI in the 1990s, long before current computing power and language models made commercial deployment practical, and now works where financial market infrastructure, technology and operations meet.

Tom describes four levels of AI productivity. The first helps an individual complete a task. The second turns successful personal patterns into tools that other employees can use. The third places agents inside production workflows while a person verifies the result. The fourth allows agents to operate without constant human review, but only within a controlled environment with monitoring, policies and evidence that the system is behaving as expected.

One of the strongest lessons from our conversation is that AI can force an organization to examine a process it may have accepted for years. Tom says Broadridge removed 30 percent of the email traffic entering one operational group after finding duplicates, messages that required no response and work that should not have existed. His own estimate is that around half the benefit of an AI program may come from simplifying the current process before advanced models are applied. That is an important counterweight to the assumption that every inefficiency needs a language model.

We also discuss the economics of enterprise AI. Tom compares current token-cost surprises with the early cloud projects that moved existing workloads without redesigning them for a different computing model. Broadridge baselines costs, tests use cases through its operations business and monitors token consumption. He shares one example in which a team's default configuration automatically moved to a newer and costlier model, showing why model selection and configuration belong in the operating discipline around AI.

For financial institutions choosing their first agentic workflow, Tom recommends high-volume, rules-based work with known inputs and outputs. A person or an automated test should be able to verify the result. He also favors several smaller agents over one large agent responsible for an entire process because individual steps are easier to inspect, replace and stop when something goes wrong.

As models become widely available, Tom believes business advantage will increasingly depend on the proprietary data, platforms, workflows and governance surrounding them. Does your organization have the process discipline, cost visibility and control framework required to put AI into production? Listen to the episode and share your thoughts with me.

Useful Links

Executive: Tom Carey, President of Broadridge's Global Technology & Operations business

Broadridge Deploys Agentic AI at Institutional Scale Across Capital Markets and Wealth Operations (May 11, 2026)

Broadridge Joins Anthropic's Project Glasswing (June 17, 2026)

GenAI Delivering Now, Tokenization Is Next: Financial Services Enters Period of Accelerating Transformation, Landmark Broadridge Study Finds (February 25, 2026)

LTX Launches Agentic AI in BondGPT, Turning AI Insights into Trading Action (June 16, 2026)

Broadridge Invests in DeepSee, Further Harnessing Agentic AI to Transform Post-trade Operations (January 8, 2026)