Takeaways from the 2025 Harvard Tech Club Conference| Behind the Bubble: Ushering in the new Era of Technology

 Saturday was an exciting day on the HBS campus if you were part of the 2025 Tech Club annual conference. As a member of the “greater tech” ecosystem in Boston, I felt privileged to hear industry leaders and a wide range of “doers” share their insights. It was advantageous to have attended the 2024 event to see where the conversation would go on the impact of GenAI. Appropriately, the theme was cast as, “the AI revolution, uncovering the transformative shifts redefining industries today.” The following blog is intended to provide highlights and my personal takeaways from the keynote sessions and breakouts that I attended. Hopefully, this will benefit those who could not attend and reinforce some key points for those of us in the room. Many thanks to the HBS Tech Club team and the generous speakers and panelists for sharing their insights.

Opening Keynote: Tom Eggemeir, CEO Zen Desk

Tom’s down to earth presentation and tone made the auditorium feel like we were having a comfortable face-to-face chat about the impact of AI on the customer service sector. Truly Zendesk has fully embraced all aspects of AI and is bring their customers along while meeting them in their comfort zone. 30% of their bookings are related to AI and he considers this the biggest revolutions since e-Commerce.
Tom’s slides told the story of adoption, but the talk spoke to understanding the demographics and human reactions behind it. In their recent survey of 50K respondents, it became clear that age is a huge factor with people under the age of 40 being most positive about these applications for AI and over 60 being the most skeptical (no surprise!). Behind the data, customer interactions are seeing less friction which translates into stronger resolutions. Banking on this, Zendesk is implementing an aggressive business model driven by resolutions.

In terms of human agents, making them friendlier, authentic, casual in tone, enthusiastic and knowledgeable increases consumer confidence. Adding personalization creates yet another key dimension for success. He gave a great example of a private jet company using personalization to anticipate and recommend leisure activities and resources to a client that had previously booked a trip to the Master’s.

There is no doubt about the cost savings of doing this right. CSX spend for human labor costing $5-$15 vs. $1.49 in the automated realm. Tom does not believe this is a bubble. It’s here to stay and in 3 years, nearly 100% of customer service will be in the hands of AI other than for VIP’s. In his estimation, many of the companies are underestimating just how fast this will happen. Hold on to your hats!
 
Nicholas Harris, CEO Lightmatter and Professor Andy Wu – Fireside Chat

Lightmatter’s Harris opened with a riddle. What did Alexander Graham Bell consider his greatest invention? It wasn’t the telephone! It was the photo phone – obviously ahead of its time. With a $4.4B valuation, Lightmatter is blowing up the chip landscape using light vs. electricity (for example) to move data between chips. Nicholas did an amazing job demystifying this highly scientific physics lesson into something that made sense to the mere mortals (like me). He articulated the many advantages of “innovating at the speed of light” from the ability to send many signals with different colors to the higher bandwidth nature of optical wires. The end result is the more efficient movement of data.

Even more fascinating were his views on AI vs. Quantum. Quite the contrarian, he does not see a clear future in Quantum Computing as the writing of programs for Quantum is hard. He proports that AI can write the algorithms and create the same benefits with a broader based of engineering support.
When asked about where the opportunities are in the chip sector – he suggests that innovation in hardware is falling behind. He gave a shout out to the engineers to focus on this area. Prompted by the amazing interviewer who asked what his boldest, way-out, predictions are – he visualized data centers on MARS or perhaps in Antarctica!

Ed Tech Panel Discussion: Mary Strain, AWS ; Siya Raj Pirohit, OpenAI, Marta McAlister, Google, Joe Bennett, MBS Moderator

Two of this dynamic panel, Mary and Marta began their careers as teachers in the Bronx. Their observations of students – using technology the wrong way or not at all – or pure fear of completing assignments incorrectly – drove them into their current roles. Fast forward to the application of AI, what is unfolding is a complete redefinition of the roles of key players in education.

Impact to the Educational Sector:
Each of the panelists are focused on different parts of the educational spectrum, but all had in common the goal of more deeply engaging their target audiences from early childhood through higher ed. A common concern is to find ways to strengthen critical thinking. Whereas AI (like “search’) has made some things easier to do, putting in “effort” to achieve an end goal shows progress. Moving forward, the skills that will be needed are critical thinking and judgement – the same that will be required in the workforce and will not be provided by AI.

OpenAI has a relatively small number of its team devoted to education. They are scrappy and entrepreneurial. Their focus is on visualizing the output to help up-level all the AI products by spending time with experts in the field – noting that “teachers are the best prompters.”
Acknowledging that only 60% of the population actually goes to college, AWS is focused on creating architectures that create lifelong learning opportunities. A likely outcome is more decentralized learning. In higher ed, AI will challenge each university’s value proposition. If you are paying for a top-notch university, stakeholders will demand that the “end product” measures up – placing more urgency into the discussion of competency-based assessment.

Evolving Skills, Tools, and Roles:

OpenAI (through ChatGBT) is also focused on providing tools to make it easier for the influencers (aka professors) – so they have recently added 800 help numbers and a WhatsApp number to provide help to classroom superusers. The panel agreed that the role of the professors will have to shift considerably as more knowledge can be acquired through AI. Professors may be come curators or assessors – forcing a massive change in the way education is managed.

With AI replacing or enhancing many of the mechanical learning tasks, the superpowers of the future will be empathy, curiosity, and the ability to learn and be “coachable.” Additionally, be able to lead a team, inspire, and connect with others. In other words, the “soft skills” of the past will be the critical skills of the future. Other differentiating factors will be: domain expertise and the ability to solve real world problems.
Among the bold predications for the future – personalized tutoring stands to become a huge opportunity as it’s considered the “holy grail” of education. But overall, the panel agreed that overall, the fundamentals of pedagogy will not change and what we should be hopeful for is reinforcement of good, teaching best practices.

Health Tech Panel Discussion: Jack Sallay, Amazon; Derek Streat, DexCare; Jeff Liu, Assort Health; Chris Fank, Hippocratic AI; Moderator, Amna Hashni, MBA

Sector Focus:
Each panelist opened by citing their unique value proposition for AI in healthcare through the prism of a 2 x2 quadrant – Clinical, Administrative, Consumer, and Provider. The first panelist from Amazon believed all four quadrants could be impacted but consumer, non-clinical is where he sees a huge opportunity to unravel and clarify cost and services for health care procedures. DexCare (a Kaiser partner) helps providers become more efficient (founded inside of Providence), Assort Health is focused on AI agents for the health care practice with a focus on safety; Hippocratic is focused on closing the gap on the worldwide shortage of healthcare professional.

Adoption:
The sentiment was that the overall category was in the early majority phase of adoption with providers further along but pharma, life sciences, and the payer space lagging behind due to compliance and other factors. One of the panelists put it in the context that the health care system as a whole lean into being skeptical as the downside risks (loss of human life) are much higher. The non-clinical side, however, (payers) need to move up the adoption curve – noting that 11K people a day age into Medicare!
Human Interaction in Healthcare:

The question was raised as to whether there would always be human interaction in healthcare. The answer was “yes” but the job of the human will change considerably. With the scarcity of nursing and other health care staff, more information will be shared through vertical AI agents which are working from strong foundational models.

Other trends and models will include:
• The emergence of “pay-buyers”
• Return power to private practices (like dentists)

Overall, the problem that needs to be solved is to create a more sustainable, transparent health care system and one that reduces the cost to patients. Patients are consumers and have choices and there will be more competition for these customers.

Pre-Lunch Fireside- Linda Sheng, GM, Minimax

Minimax AI is an advanced AI-powered video creation tool that transforms text into dynamic video sequences with precise control over composition and camera movements. The CEO is an ambassador for AI and believes her role is to showcase what AI can do as a positive force.

This discussion centered around the overall threat/opportunity of AI and the role humans will play. The CEO clearly believes that a successful founder must be AI-equipped and somewhat technical but does not see a clear threat or “red button” that would blow up humanity.

Use cases will drive AI success. For example, the largest freelance network, FIVR adopted AI to generate IP protected voices of their freelancers to save the actors from having to repeat voiceovers over and over. Now they get paid for use of their voice more frequently and more efficiently.
Speaking to AI “morality,” Minimax is highly sensitive to content. Areas such as local customs and religion and social “red lines” are closely watched. Another key area of focus is “access.” Making sure that great creators/superusers can use AI for free if they are making important contributions. The Open-Source model provides this access and its important to support freedom of code.

NLP and Generative AI Panel: Aravind Suresh, OpenAI, Shashank Chaudhary, Google Gemini; Uriel Kejsefman, Duolingo; Avijit Ghosh, Hugging Face

The opening discussion focused on the unique challenges of AI – in particular the building blocks and infrastructure precedes the product/applications. Duolingo’s product lead provided an exciting use-case on all the ways that AI had enhanced their offering – creating entirely new ways of teaching, content development, and impact on productivity through use of new tools such as notetaking. Role playing, transactional writing, and features such as “explain my answer” are all part of the new level of offerings. It was also noted that AI can help create many more exercises, and courses and help modularize the curriculum.

The discussion broadened to talk about the socio-economic impact of AI. One of the areas of consideration is making sure that small companies have access- pointing out that that’s one of the advantages of DeepSeek.

There was also an interesting talk track pointing out that not all AI will come from the same point of view. For example, the group has seen analysis that demonstrates bias against the US. It can also be biased for another region or country – reflecting the local “mindset” foreign policy or other national biases negating neutral guidance or responses. The bottom line of this discussion was encouraging US companies to invest in US-based models reflecting American value systems.
Evolution of the PM role:

The Product Manager has to uplevel skills to include:
• Prompting
• Creating a product that works but still needs humans
• Develop and socialize data strategy because bad data is devastating
• Break down outputs

And continue to enforce product/market Fit and A/B Testing which is still relevant.

AI Tools for Enterprise: Austin Johnson, Zapier; Nic Moralis, Cohere; Drew Bent, Anthropic; Tejas Pathak, Scale AI; Alisha Bhanji, MBA

Most of the panels and keynotes were superusers of AI, but reflecting on the general climate was a bit more sobering – especially as AI investment is growing inside the enterprise. Noted by one of the panelists, today’s environment dictates that we show ROI one quarter after Capex. Therefore, CIOs are getting disillusioned with AI as they do not see the usage of GenAI tools increasing at a rate that justifies the level of investment. The major call to action here is to move from a top-down implementation to a bottom-up adoption model.

Fluency in AI:
Delegation (automation of tasks); Description; Discernment (understanding outputs); and Diligence (responsible use of AI) were suggested as the four key pillars to achieve fluency. One of the panelists brought up a point of view – “Do unto others (in AI) as you would have others do to you,” using employee evaluations as an example of something you may not want to delegate to AI!

Use Cases & Best Practices:

Cohere discussed business value using an enterprise database software company as an example where 200 AI features have been implemented across its applications. With data at the center of AI’s validity – citing sources is more critical than ever.

The further discussion on data was to respect the existing rules on data. For example, asking the question – would my customers be happy with where the data is going? Generally speaking, security, compliance, re-thinking workflows and workforce dynamics vis-à-vis AI agents is still evolving. Additionally, different size models and pricing will evolve to accommodate different size entities and usage of the Cloud.
Looking ahead, AI knowledge, data gathering, and analysis has to have parameters – it cannot just be “let loose.” One speaker used the analogy – we just invented single brain cells, now we have to string them all together.

Business Focused Roles Panel: Maggie Yang, Amazon; Fadzi Makanda, Google; Jannis Woelk, Rippling, Cindy Fan MBA, Moderator

This panel was heavily focused on helping the student audience define non-technical roles they can seek out in tech with an emphasis on AI. Two of the panelists represented corporate development and one was in a business operations role.
The corporate development roles required top notch research on trends and the ability to make a clear case for expanding into new areas – working closely with product managers, engineers, and other key stakeholders across the company. Each faced a different set of challenges. These included:

• Build vs. Buy
• Identifying differentiators in highly competitive markets
• Defining and meeting success metrics

Key skills for these roles are varied, but being able to bring perspective/judgement; gaining the trust of peers cross-functionally; bringing high quality ideas to the table; listening and having empathy are needed alongside good, solid general knowledge of the technology.

Closing Keynote: Diego Lomanto, CMO Writer

The final keynote appropriately represented a company that addresses the full stack of Generative AI. As such, Writer has enjoyed success shepherding their customers on the journey by identifying “the most miserable parts of the job” and using AI to automate them.
The full stack, in Writer’s terms addresses: The brain, context, control, workflow, and impacts (see slide). Diego presented some key findings from the Adore.me use case to demonstrate how AI was used at each interval to improve the experience of the product managers and the end users. Two key tasks truly stood out of this overview. One involved using AI to write thousands of product descriptions – an excruciating task for a retail product manager. The second involved mining customer 200,000 insights from across all media videos, case studies, interviews, testimonials, etc. to create compelling content using the Knowledge Graph. Finally, he addressed Agentic AI – applied to an exploding inbox. When deployed well, an agent can understand content and context in the email so that responses are tailored and appropriate to the severity of the situation.

Companies are finding it hard to put all the pieces together and there’s clearly a role for intermediaries, full-service providers, and partners to act as catalysts and consultants to guide adoption. My takeaway for the day is that everyone one of us has a chance to be an AI ambassador. It’s clearly as big as any phase of the internet roll-out. AI emissaries are needed in each sector, at every level, and across all functions of business. Being proactive will ensure that we uphold business ethics and keep the technology accessible to all types of businesses.

[For the record, this blog post was NOT AI generated – you had to be there].