Tech Talks Daily

What happens when an AI agent begins influencing business decisions without fully understanding the systems, processes and dependencies behind them?
In this episode, I speak with Bert van der Zwan, CEO of Bizzdesign, about the gap between enterprise AI expectations and the results many companies are seeing in practice. Bert has spent more than 25 years in software and SaaS leadership, including executive roles at Webex, Bynder, Twinfield and Unit4.
Bert offers a candid assessment of the current AI market. He believes AI will have a lasting effect on businesses and society, but argues that expectations for near-term financial returns have become inflated. Many companies are spending money on tools and experimentation without reducing costs, consolidating software or producing new revenue.
That does not mean experimentation is a mistake. Bert sees it as a necessary stage. The harder question is how companies move from a growing collection of pilots to AI capabilities that can operate dependably inside the business.
One barrier is fragmented organizational context. Large enterprises have often grown through a combination of internal expansion and acquisitions, leaving behind disconnected applications, inconsistent data definitions and processes that cross several departments. An AI system working with only part of that picture may make a fast decision, but that does not make it a good decision.
Bert argues that AI needs an authoritative view of how the enterprise works. Systems, processes, ownership, dependencies, approval status and policy restrictions must be visible and consistently defined. Without that shared context, AI may reproduce existing silos or make them worse.
We also discuss the risks boards and technology leaders should consider as AI agents become involved in operational decisions. These include unreliable data, unclear accountability, legal exposure, weak governance and an incomplete view of the process being changed. Human oversight remains necessary, particularly when an automated decision could affect customers, employees or major investments.
Bert then introduces the idea of "bespoke from the cloud." Traditional SaaS products were built around largely standardized interfaces and workflows. AI-assisted development could make software far easier to personalize around individual customers and use cases. This may give users greater control, but it could also challenge long-term software contracts and the economics that have supported the SaaS market.
For leaders trying to connect AI spending with business results, Bert recommends beginning with visibility and a clearly defined outcome. Every initiative should be judged by whether it reduces costs, increases revenue or shortens the time required to deliver value.
If AI depends on understanding how a company actually works, have businesses invested enough in creating that shared understanding before adding agents to their operations? Listen to the episode and share your thoughts with me.

Can you still trust an incoming phone call when AI can imitate a familiar voice, personalize the conversation and target information specifically to you?
In this episode, I speak with Alex Quilici, CEO of YouMail, about how artificial intelligence is changing phone fraud and why the personal devices carried by employees are becoming part of the corporate attack surface.
Alex explains how YouMail uses data from its consumer call-protection service to identify scam behavior, understand the type of fraud taking place and connect those patterns with the phone numbers involved. Advances in large language models have improved this analysis, but the same technology is also helping criminals build far more convincing campaigns.
Generic robocalls are being replaced by personalized conversations designed to extract information, impersonate trusted people and manipulate victims. Fraudsters can use AI throughout the attack chain, from identifying targets and analyzing stolen data to generating dialogue and adapting an approach during the call. Alex argues that attackers have adopted these capabilities faster than many defenders expected because successful fraud produces an immediate financial return.
The conversation also examines why voice biometrics can no longer be treated as sufficient proof of identity. As voice-cloning tools improve, companies may need to combine multiple forms of authentication and move sensitive communications into trusted applications. A call received through a banking app, for example, could give the customer greater confidence that the caller really represents their bank.
For businesses, the risk extends beyond company-managed technology. Attackers can identify where someone works, learn about their role and contact them through a personal phone that may sit outside corporate monitoring. An employee's private number can therefore provide another route into the business through impersonation and social engineering.
Alex also makes a persuasive case for collecting less personal data. Personalization can improve a service, but every additional piece of information becomes something an attacker might obtain during a breach. His advice is to identify the minimum information needed to deliver the intended experience rather than gathering data simply because it may prove useful later.
Despite the seriousness of the threat, Alex offers evidence that coordinated action can produce results. He has seen brand-impersonation campaigns reduced from tens of millions of calls each month to around 100,000 through monitoring, disruption and cooperation between businesses and telecommunications providers.
If AI is making fraudulent calls harder to recognize, should businesses stop treating the telephone network as a trusted communication channel by default? Listen to the episode and share your thoughts with me.

Why do AI agents and applications look impressive in demos but struggle when companies try to deploy them in production?
In this episode of Tech Talks Daily, I speak with Nikunj Bajaj, co-founder and CEO of TrueFoundry, about why enterprise AI has become a systems problem, what companies need to move AI from proof of concept to production, and how better infrastructure can improve reliability, governance, security, observability, and cost control.
Before founding TrueFoundry, Nikunj worked at Meta on conversational AI systems serving more than a billion users and contributed to the company's internal machine learning platforms. He explains how developers at Meta could concentrate on solving business problems while infrastructure handled logging, monitoring, deployment, and governance by default. In many enterprises, the same journey from an AI idea to a production application can still take weeks or months.
Nikunj argues that increasingly capable AI models are not necessarily the biggest barrier to enterprise adoption. The harder challenge is building reliable systems around them. Companies need to know what happens when a model becomes unavailable, how an agent is behaving, which data it can access, how much it is costing, when a human should intervene, and whether there is a kill switch when something goes wrong.
We discuss why AI proofs of concept often fail when exposed to real users. Controlled demonstrations rarely reproduce production conditions such as unexpected prompts, malicious actors, heavy workloads, model outages, latency, and dependencies between multiple components. Even when individual parts of a system perform reliably, combining them can create failure rates that businesses cannot accept for mission-critical workflows.
The conversation also examines the infrastructure required as companies introduce multiple AI models and agents. Nikunj explains the roles of model gateways, MCP gateways, and agent gateways, and how bringing these components together through an AI gateway can give enterprises a control plane for observing and governing AI traffic.
Cost is another major challenge. Nikunj explains why sending every request to the most powerful model can waste significant amounts of money when smaller or cheaper models could produce comparable results for simpler tasks. Intelligent model routing can help companies balance quality, latency, availability, and price. He shares how organizations using this approach have reduced model costs by as much as 75 to 80 percent in some production environments.
We also discuss what reliable multi-agent systems require in practice. Companies need clearly defined boundaries for what agents can do, escalation routes to other agents or people, safeguards against infinite agent loops, and complete audit trails of interactions and decisions.
For CIOs, CTOs, AI engineering teams, platform leaders, and companies trying to move generative AI and agentic AI into production, this conversation provides a practical guide to the infrastructure decisions that determine whether AI applications remain impressive prototypes or become reliable business systems.
The next stage of enterprise AI will not be defined by models alone. Companies that can connect, observe, govern, secure, and control their AI applications while managing costs will be better positioned to turn experimentation into dependable production systems.

What if the biggest barrier to better customer service isn't how quickly employees work, but how much time they lose coordinating with everyone else?
In this episode of Tech Talks Daily, I speak with Kevin Yang, Head of AI at Front, about why customer conversations are becoming a valuable source of business intelligence, how AI can improve work across entire teams rather than simply making individuals faster, and the hidden coordination costs affecting customer operations.
Kevin brings a unique perspective to the conversation. Before joining Front following its acquisition of his AI voice-of-customer company, Syllable, he spent 15 years as an entrepreneur. While building an office food delivery business, he experienced firsthand how customer conversations could reveal problems that traditional surveys and dashboards failed to identify. By analyzing customer feedback at scale, his team could connect specific issues directly to retention, account growth, and referrals.
Today, AI makes it possible for companies to analyze enormous volumes of customer conversations and turn unstructured feedback into intelligence that can inform decisions across product development, sales, marketing, and customer success.
Kevin shares how Front analyzes conversations to understand why deals are lost, why customers leave, and which topics are associated with higher sales conversion rates. The result is a feedback loop that helps companies direct product investment toward problems customers genuinely care about while giving sales and marketing teams a clearer understanding of the conversations that influence buying decisions.
But the episode also challenges the assumption that giving every employee an AI assistant will transform productivity.
Front's Coordination Tax research found that teams can spend almost three hours coordinating work for every hour spent solving customer problems. When a single customer request requires input from sales, finance, support, operations, or external systems, employees can lose time to emails, Slack messages, meetings, handoffs, and information searches.
Kevin explains why making one person faster does little to solve this problem if the rest of the workflow remains fragmented. The bigger opportunity is to use AI across end-to-end processes, automatically handling research and analysis while allowing people to concentrate on work requiring judgment, empathy, relationships, and human decision-making.
We also discuss the growing use of AI agents in customer operations and why governance becomes harder as companies move from experimenting with one agent to managing many. Kevin outlines the need to measure whether agents follow processes correctly, understand customer satisfaction, identify where failures occur, and continuously improve the knowledge and guidance available to AI systems.
For business and technology leaders considering where to apply AI, Kevin offers a practical starting point. Map the work your teams perform into three categories: tasks AI can automate, tasks AI can support with human review, and tasks that should remain human. This helps companies focus investment where AI performs well rather than forcing automation into customer interactions that depend on empathy, context, and relationships.
For anyone responsible for customer experience, AI strategy, operations, or digital transformation, this conversation provides practical ideas for turning customer conversations into business intelligence, reducing coordination friction, designing better workflows, and introducing AI agents with greater visibility and oversight.
The opportunity is not simply to make individuals work faster. It is to redesign how work moves across the organization so employees spend less time coordinating and more time solving the problems that matter to customers.

What happens when your next customer is represented by an AI agent that can research products, compare prices, evaluate suppliers, negotiate terms, and make purchasing decisions?
In this episode of Tech Talks Daily, I speak with Ian Kahn, Partner and Customer and Commercial Excellence Platform Leader at PwC, about the rise of the Intelligent Customer Edge and why companies need to rethink how they sell, market, price, serve customers, and compete as artificial intelligence changes the buying process.
Much of the enterprise AI conversation has focused on helping employees become more productive. Ian argues that this overlooks a much bigger change already taking place. Customers are using AI to research products, compare alternatives, evaluate pricing, and make decisions. In some consumer and business markets, AI agents are already being given permission to make routine purchases.
Companies are no longer selling only to people. They increasingly need to serve customers whose AI agents expect accurate product information, transparent pricing, availability, service history, and performance data that can be discovered, verified, and understood by machines.
This creates a serious problem for companies operating with fragmented front offices.
Marketing, sales, pricing, commerce, and customer service have traditionally operated as separate functions, each with its own technology, data, processes, incentives, and performance measures. Customers do not experience companies through those internal structures. They expect consistent information and relevant experiences across the entire relationship.
Ian explains why adding AI to each department independently will not solve this problem. Companies risk making existing processes faster without improving the customer experience or business performance. Instead, he argues that leaders need to reconsider the operating model behind the entire customer journey.
The Intelligent Customer Edge is PwC's approach to bringing these commercial functions together into a connected system centered on the customer. Powered by proprietary company data and AI, the system can continuously learn from customer interactions, support real-time decisions, and help companies respond to changing customer needs.
We also discuss the idea of the commercial brain and why proprietary data could become one of the most valuable competitive advantages available to companies adopting AI.
Most businesses already possess customer records, transaction histories, operational information, market signals, service interactions, and other data their competitors cannot access. Yet much of that information remains fragmented across systems and departments.
Ian explains how connecting these sources can create an intelligence layer that informs pricing decisions, marketing activity, sales opportunities, service interactions, and the moments that matter throughout the customer relationship.
For CEOs, chief customer officers, marketing leaders, sales executives, CIOs, and technology teams, the conversation offers an important lesson about AI transformation. The companies achieving meaningful results are not starting with the technology. They begin with customer outcomes and redesign the work, decisions, workflows, and operating models required to achieve them.
Human judgment remains an important part of that model. AI can process large amounts of information, identify patterns, provide recommendations, and handle routine tasks consistently. People continue to bring judgment, creativity, empathy, relationship-building, and strategic decision-making to customer interactions where trust and context matter.
Ian argues that the goal is not to choose between people and AI. Companies need to design customer systems that use the strengths of both, determining where automation can improve speed and consistency and where people can create greater customer and commercial value.
Trust, governance, explainability, and accountability also become more important as AI agents are given greater authority. Rather than treating guardrails as barriers to adoption, Ian explains why companies should design controls into AI-enabled customer processes from the beginning.
The conversation also examines the cost of waiting. Customers are already adopting AI, and businesses that continue relying on fragmented front-office operations risk falling behind competitors capable of responding faster, providing better information, and creating more relevant customer experiences.
Ian offers practical advice for companies deciding where to begin. Start with the customer journey. Understand how customer behavior is changing, identify where friction exists, determine how AI could improve the experience, and establish clear measures for customer outcomes and business value before investing heavily in new technology.
For business and technology leaders under pressure to deliver growth, improve margins, control costs, and demonstrate returns from AI investment, this conversation provides a practical framework for redesigning the front office, using proprietary data more effectively, preparing for AI agents as buyers, and creating better customer experiences.
Your customers are already using AI. Some AI agents are already making purchasing decisions. The question for companies is whether their customer systems, data, commercial models, and operating structures are ready to compete for business when the buyer on the other side of the transaction is no longer always human.