AI Today: Make AI earn its keep.

At TechBreakfast, the innovation arrived with breakfast. Kathleen Walch and Ron Schmelzer worked on events where startups demonstrated their latest ideas. Around 2016, Walch later recalled, AI began appearing throughout those pitches. Voice assistants were the familiar examples. The demos made the technology look approachable. Getting it to work inside an organization was another matter. That distance between the presentation and the project became their business opportunity.

The demo was only the beginning

Walch and Schmelzer founded Cognilytica in 2017. AI Today began that year, too. Their company combined AI research, education, and advisory work; the podcast let the same questions travel beyond a consulting engagement. What were organizations actually doing? Which problems justified AI? What happened when the promising experiment met a department with its own rules?

The founders brought different training to the microphone. Walch studied marketing at Loyola University. Schmelzer studied computer science and electrical engineering at MIT and earned an MBA at Johns Hopkins. One useful way to read their partnership is as a bridge between the buyer’s question and the builder’s question. A working AI project has to satisfy both.

A microphone with a job to do

AI Today’s product is explanation. Its archive includes interviews, an AI glossary, prompt engineering discussions, and industry use cases. Its stated audience includes enterprises and public-sector agencies. The listener may be a project manager translating technical language for colleagues, or a business leader deciding what to try. You can listen without buying a model, installing software, or becoming a data scientist.

That places it in a crowded audio market alongside shows such as Practical AI, The AI in Business Podcast, and NVIDIA’s AI Podcast. AI Today’s useful editorial emphasis is implementation: the decisions surrounding the technology. Under PMI, project leadership has become particularly prominent. The current official page names Walch as host, while the earlier Cognilytica archive foregrounded both founders.

The business has more than one entry point. Listening is free. Cognilytica’s broader offering included paid training and certification, now folded into PMI’s portfolio. The current podcast also carries sponsorship: its September 30, 2026 interview with Samia Waqar names SnS Coaching Consulting Services as sponsor. Free access and commercial education coexist here. The listener can learn a little; the professional can pursue a credential. Those are different commitments. Hearing an interview asks for attention; following a formal curriculum asks for study and assessment. The pairing gives the business a way to explain a problem publicly while offering professionals a more structured route through it.

The warehouse says no

In a March 2025 Projectified interview, the founders described a wonderfully unglamorous obstacle. A team assumes it can obtain information from the company’s data warehouse. The people controlling the warehouse refuse. Months pass. The project has a technical ambition, but it lacks access to its essential ingredient. A more impressive model cannot negotiate that permission.

Walch’s explanation of their methodology’s origin centers on this kind of sequencing mistake. Teams jumped ahead, then discovered work they needed to do earlier. The founders’ diagnosis also includes poor data quality and weak business justification. Their advice permits a surprisingly plain answer: sometimes ordinary automation, or a person, is the better way to solve the problem.

There is a cost lesson in that permission dispute. AI work consumes time, money, and people before it delivers a return. A useful budget must consider those resources, not merely the tool subscription. The podcast can help a listener frame that calculation. It cannot turn an inaccessible dataset into an accessible one.

Six phases, plenty of return trips

Cognilytica’s response was CPMAI, a methodology now offered through PMI. It begins with business understanding, then examines and prepares data before model development, evaluation, and operationalization. The ordering puts the customer’s need ahead of the tool choice. The phases are iterative: evidence can send a team backward to revise its assumptions.

For someone buying the related PMI-CPMAI offering, the product goes beyond listening: a preparation course and a certification assessment. PMI lists a 120-question exam lasting 160 minutes. Fees depend on region and membership. This is education in managing AI work. It complements the technical people who build a system and the business people who decide whether its results are worth using.

Why PMI wanted the microphone

On September 19, 2024, PMI acquired Cognilytica. The purchase brought training, certification, research, and AI Today into an organization already serving project professionals. PMI’s announcement described Cognilytica’s experience with hundreds of real-world AI projects. For the founders, the appeal was distribution and institutional credibility. A small specialist’s teaching could reach a broader professional community.

“Think big. Start small. Iterate often.”Kathleen Walch / PMI interview, 2024

The podcast’s subsequent guests show that direction. Lee Lambert, a founder of the PMP certification, discussed communication, leadership, and adaptability in an episode recorded at PMI Global Summit 2025. In July 2026, Walch marked the show’s tenth season by revisiting recurring questions about data, governance, and expectations. The tools had changed. The organizational questions kept returning.

Bring one question to Monday

The thing to copy is small enough to fit on a meeting agenda: what result would make this project worth doing? Write that down before choosing a model. Then check whether the required data exists and whether the team can use it. AI Today gives the curious listener vocabulary for that discussion and examples of the people having it.

Its value depends on what happens after listening. An organization still needs technical expertise, an owner for the decisions, and people willing to review results. A podcast cannot supply those conditions. But it can make a familiar meeting more productive. Before someone unveils the next dazzling demo, somebody at the table can ask where the data will come from.

Keep the conversation going