Building a Trusted AI Voice With Voices

What good is a new treatment if patients still struggle to get the approvals, appointments, and support needed to receive it? In this episode of Tech Talks Daily, I speak with Amir Sharif, president and co-founder of Leap, about the operational infrastructure behind specialty care. While AI in drug discovery attracts attention, the journey from a prescription to treatment often depends on disconnected organizations, incomplete information, and manual coordination. Amir describes how Leap combines drug sourcing, distribution, care coordination, and employer benefit arrangements to address some of those gaps. His interest in patient experience grew through work helping build healthcare businesses including Care/of and Tend. Those experiences led him to question why everyday expectations around scheduling, pricing, and customer support can be so difficult to meet in healthcare. Specialty therapies bring another level of complexity because they can require clinical administration, recurring appointments, insurance authorization, and expensive medication procurement. We look at what happens after a physician writes a prescription. Approval may depend on the patient's health plan, supporting medical information, an administering provider, and a permitted site of care. Coordinating those pieces can fall to the patient, even while they are dealing with an illness. Amir explains why access depends on the administrative process as well as the treatment itself. Drug pricing provides another part of the picture. Under a buy and bill arrangement, a provider purchases a drug, administers it, and bills for it. Amir says the margin added to the acquisition cost can range from 40 percent to 400 percent. These are his reported figures rather than a finding independently established in this episode. He describes Leap's model as passing its drug acquisition cost to the employer or plan sponsor and coordinating administration through home infusion or partner clinical sites, depending on patient preference and clinical requirements. The technology discussion becomes particularly practical when Amir describes the data behind identifying eligible patients. Leap receives feeds containing diagnosis codes, claims, and prior authorization information. Its data engine normalizes those feeds so the team can act on signals that a patient is receiving, or may soon be prescribed, a therapy within the program's scope. This is a care coordination process. Amir repeatedly says Leap does not replace the prescriber or take control of clinical treatment decisions. He also describes an average patient encounter involving over 20 coordination touchpoints. Many external parties still communicate by email or fax rather than through software interfaces. Leap uses agentic workflows to support appointment scheduling, shipment tracking, and medical order processing. According to Amir, those workflows remain supervised by care teams, and patients speak with a person when they contact the service. For business leaders assessing AI returns, his account of model costs is especially useful. Some workflows the team considered two quarters earlier would have cost tens or hundreds of dollars per run in token consumption, he says. As models become cheaper and more capable, different operational uses become viable. The interview does not provide a measured before and after cost comparison, so the lesson is about evaluating changing economics rather than declaring a proven return for every workflow. Leap reports a patient Net Promoter Score of 92, with surveys sent after every encounter. Amir describes this as evidence of patient satisfaction, alongside feedback gathered by care teams. Satisfaction is useful, but it should not be confused with proof of improved clinical outcomes or faster treatment across every patient. Employers can request an impact analysis comparing historical claims with the proposed cost of Leap's service. Where would you begin when assessing specialty

What happens to a good idea when the person who understands the problem best cannot express it confidently in the language used by the rest of the team? In this episode of Tech Talks Daily, I speak with Gonçalo Gaiolas, Chief Product Officer at DeepL, about real time voice translation and the work required to make a multilingual conversation feel natural. Our discussion moves from the business cost of language barriers to the technical choices behind speaker identity, tone, speed, and the handling of sensitive meeting data. Gonçalo begins with a cost that is difficult to place on a spreadsheet. An employee may understand a local market, a product, or a customer need but struggle to contribute in the company's shared language. The organization can miss that knowledge even when the person is present in the meeting. Customers also tell DeepL that language barriers affect the time needed to make decisions and understand nuance. These are experiences he describes, rather than independently measured results presented in the interview. Live translation introduces a demanding set of constraints. People pause, interrupt, change direction, and build meaning at different speeds across languages. A system needs enough context to produce an accurate translation while responding quickly enough to remain useful. Waiting can interrupt the flow, but responding too early can lose the meaning of the sentence. Tone creates another challenge because urgency, hesitation, and a question can carry information that the translated words alone do not capture. We discuss DeepL's approach to retaining the characteristics of individual speakers. Gonçalo describes separating speakers and carrying elements of voice, rhythm, and emotion into the translated audio. He also explains that some original audio remains audible at a lower level. His aim is to help listeners follow who is speaking while understanding the translated conversation. He says users have options around retaining vocal identity, acknowledging that people may feel differently about using that capability. The enterprise adoption discussion is equally practical. Gonçalo describes a desktop app intended to provide a consistent experience across Zoom, Microsoft Teams, and Google Meet. A global workforce may use one platform internally and encounter several others when speaking with customers and partners. For him, working across those environments is part of the product's usefulness, alongside the quality of the translation itself. Accuracy remains a recurring theme. Gonçalo describes testing with real users and evaluating whether the result works in the customer's context. He gives the example of an engineering department whose acronyms and specialist terms can confuse a model. DeepL's customization capabilities, he says, let customers supply terminology and guidance about how language should be handled. The example offers a useful reminder that a fluent sentence can still misinterpret the way a particular team speaks. We also discuss procurement and governance. Meeting audio can contain sensitive business information, so buyers need to understand technical architecture, policies, and how their data is processed. Gonçalo says DeepL does not use customer data to train its own models. The interview does not supply the contractual scope, retention periods, or detailed controls behind that statement. Those remain questions for buyers to check against the service and agreement they intend to use. Examples range from conversations between parents and teachers to contact centers, large events, and use alongside human interpreters. Gonçalo describes customer demand that extends beyond the scenarios the team originally expected. I also ask about translating podcasts, which leads to a discussion of possible uses outside live meetings. We keep that possibility separate from a confirmed podcast production service or a promise about this show's availability in other languages. I'd welcome your though

What if the most useful industrial robots are the ones helping experienced people continue working in safer conditions? Recording from the show floor at IFS Unleashed 2026 in Orlando, I spoke with David Askey, CEO of Ascend Robotics, about construction and maintenance machines working in spaces that are difficult or dangerous for people to reach. We discussed the gap between a robotics demonstration and equipment that can perform reliably at a customer site, where conditions rarely match the original plans. David shares a ship painting example in which he reports work completed three to four times faster than with typical human crews, alongside approximately twenty five percent material savings. He also describes more consistent coating thickness and explains why people continue to monitor and direct the work. A separate example involves cleaning and coating a factory ceiling during a data center conversion, where live electrical infrastructure added to the difficulty of the job. One of the most interesting parts of our conversation concerns the people operating these machines. David describes hiring experienced workers from painting crews as robot operators and working with unions as the equipment enters the market. For workers affected by the physical demands of construction, operating a robot can provide a way to keep contributing their knowledge from a different working environment. We also discuss how robots adapt to buildings as they actually exist and how operators help them decide what to paint and what to avoid. David explains how AI tools have accelerated his team's software development and why connecting the data collected by robots to enterprise systems is a separate challenge. He describes IFS as a bridge between work happening in the field and the information management teams need. For anyone considering robotics, the conversation offers practical questions about reliability, safety, integration and the outcomes worth measuring. Where could robots support your workforce, and what would you need to see before putting them to work? Please share your thoughts with me.

What happens when individual AI agents begin working together as digital teams, and how much more can businesses achieve when automation moves beyond a single task? Recording live at IFS Unleashed 2026 in Orlando, Neil C. Hughes reconnects with Somya Kapoor, CEO and Co-founder of IFS Loops, to discuss how digital workers have developed since their previous conversation in New York. Last year, the focus was on individual digital workers performing specific tasks. Now, IFS is bringing several workers together to support wider business processes, with early customer deployments providing evidence of what these systems can achieve. Somya explains why the initial approach centered on low-risk, high-impact workflows rather than attempting to automate complex processes all at once. She shares how some customers who began with a single digital worker have expanded to two or three, creating opportunities to automate more work across procurement, asset management, and field service. The conversation also examines customer examples from IFS Unleashed. Somya discusses KLM's reported efficiency improvements, the potential annual savings associated with invoice processing at Eight Group, and how Kodiak Gas is using a digital worker to make spare-parts identification and purchasing significantly faster. These examples provide an interesting perspective on enterprise AI at a time when many organizations are still struggling to move beyond pilot projects. Rather than beginning with large transformation programs, Somya describes an approach built around focused deployments, measurable results, and the opportunity to expand when the initial benefits become clear. But the technology is only part of the challenge. Somya also explains why change management can become a greater obstacle than building the agents themselves, particularly when employees and business leaders are unfamiliar with allowing digital workers to perform operational tasks. As AI agents become increasingly capable of working together, will organizations find greater success by starting small and building confidence through results? And could digital teams provide a more practical route from AI experimentation to everyday business value? Listen to the conversation and share your thoughts on where digital workers could make the greatest difference in your organization.

What if the biggest obstacle to industrial AI isn't the technology itself, but the data businesses are trying to use with it? And how much could manufacturers and industrial operators achieve if they could connect information from across the entire life of their assets? Recording live at IFS Unleashed 2026 in Orlando, Neil C. Hughes speaks with Shawn Gregerson, VP APM & MES Business Development at AVEVA, about how industrial companies are using AI, operational data, and predictive maintenance to improve reliability and make more informed business decisions. Shawn explains why industrial AI is increasingly becoming a data challenge rather than an algorithmic one. Although AI models and analytical techniques have developed significantly, many organizations still struggle with disconnected information spread across engineering, operations, maintenance, and asset management systems. The conversation examines how AVEVA is bringing these different sources together through its industrial knowledge graph technology, creating a more reliable foundation for AI applications that need to understand the relationships between assets, processes, and operational conditions. They also discuss the growing financial pressures facing industrial organizations. Shawn highlights the increasing cost of downtime and the challenge of maintaining assets that are operating beyond their original design life. With AVEVA and IFS working together on Continuous Asset Decision Intelligence, or CADi, the discussion turns to connecting asset planning, design, operations, maintenance, and optimization so that decisions can be informed by real operating conditions. Shawn also shares where industrial AI is already delivering results, including predictive maintenance systems that monitor equipment health, identify early warning signs, and help maintenance teams determine what action to take before failures occur. Other applications include predictive quality, manufacturing throughput, and the use of generative AI to make industrial information easier to access. Looking beyond individual applications, they consider how connected industrial data could support agentic workflows and enable businesses to automate more of the tasks that employees currently perform manually. As industrial organizations face pressure to improve efficiency while extending the life of existing equipment, could connecting their operational data deliver greater value than investing in increasingly sophisticated AI models alone? Listen to the conversation and share your thoughts on how industrial AI is changing maintenance, asset management, and operational decision-making.