Looking Back: Insights on Data, AI, and Analytics

As we prepare for this year’s Own the Unknown? interview series, we reflect on some of the themes repeatedly mentioned by our five thought leaders: Jonathan Reichental, Ph.D. , Kristen Kehrer , Donald Farmer , and Tom Davenport & Ian Barkin . Over the past four months, these monthly interviews spanned topics from smart cities to citizen automation. Three themes appeared throughout all four, mirroring topics that are important to Further and our clients.?

Each month, we dedicate two newsletters to our thought leaders. The first is a review of their writing and speaking. The second is an overview of our conversations with them. We encourage you to go back to past issues, or the full conversations, for more about them. This is a bonus issue celebrating the new year. Our regularly scheduled January newsletter will be posted early next week.

1. Data Governance as an Enabler, Not a Barrier

In his interview, Farmer challenged the notion that governance should be purely restrictive. “We see data governance as a protective negative capability. We go in and we stop bad things happening,” he said. He used the metaphor of “preparing a house for grandchildren”—removing fragile items so kids “can run around and be as free and as happy as they want to be.” In Farmer’s view, this proactive approach to governance expands freedom and fuels innovation, rather than stifling it.

Kehrer likewise stressed that effective governance underpins reliable, reproducible processes. She discussed how experiment tracking and data versioning help teams avoid confusion: when every model iteration is logged, it’s far easier to share and validate results. If businesses ask for fresh analysis on a prior model, Kehrer can precisely revisit the same dataset and parameters—ensuring consistency. “I can go and with data versioning, I can get that exact same dataset.”

Reichental also sees data governance as an opportunity rather than a hindrance. Drawing on his experience as Palo Alto’s former CIO, he noted that many city decisions were deferred simply because officials didn’t have quick access to the right data. “I’d sit in a council meeting on a Monday evening, and the question would come up—‘What data do we have?’ And so often the response was, ‘We don’t have it right now, but we can get it for you.’” To Reichental, strong governance is more proactive and ensures data is at one’s fingertips.

2. Culture & Collaboration

Farmer sees a critical need to keep data front and center. “They orient their decisions around data, but they’re not necessarily driven by it,” he said, describing how executives flash a dashboard at the start of a meeting yet soon drift off into anecdotal debate. He believes creating a data driven culture requires widening the participation through data literacy.

For Davenport & Barkin, collaboration also means tapping into the grassroots of an organization. Barkin observed that “You discover half the organization has built automations in secret because official IT channels were a year behind.” Such “citizen developers” or “citizen data scientists” build prototypes rapidly, then hand them off to the pros if needed. Davenport commended a hybrid approach: “That’s the best of both worlds,” he said, letting frontline innovators spark new ideas while experienced data scientists ensure large-scale reliability.

Meanwhile, Kehrer underscored that good processes—from MLOps to experiment tracking—boost alignment among data engineers, data scientists, and business stakeholders. Clear scoping and cross-functional communication can prevent the constant churn of ad hoc requests. Governance, in this sense, becomes the foundation that keeps everyone playing from the same sheet of music.

3. Generative AI

Kehrer framed large language models (LLMs) as adding complexity that has made MLOps increasingly important. In addition to longstanding issues like model drift, and hidden biases, we now have even more version control issues to worry about. It’s almost impossible to keep up without MLOps techniques. Also, she encouraged practitioners to explore their options. She sees large language models (LLMs) as pushing data science well beyond the familiar territory of tabular data. “If you’re a data scientist and you’re currently only on tabular data,” she said, “I would urge you to start looking at the tutorials, to play with LLMs or to work with image data.”

Barkin described “an absolute blossoming of generative technologies that can do so many powerful things” and predicted a sweeping democratization of AI development. Barkin explained, “It literally only depends on your ability to articulate what the need is, and it will start building it for you.” He likened generative AI to a universal translator—converting human language directly into working code or workflows. This, Barkin contends, “is creating a renaissance of entrepreneurial activity,” in which domain experts can rapidly prototype solutions that once required extensive programming or IT intervention.

Looking Ahead

As we look ahead to 2025, we’re excited to continue these conversations with thought leaders such as Matthew Lungren MD MPH , and AI ethicist Olivia Gambelin . Our next interview, with Matthew, is coming up very soon. Our January newsletter will discuss the event in more detail. You can expect the newsletter early next week, and the interview will be on January 8th, at 1 pm ET.

You can watch all past interviews here.