Presenters

Source

Beyond the Buzzword: Building AI That Actually Delivers Value 🚀

We’re living in an AI-saturated world. From our coffee machines to our toasters, it feels like AI is being added to everything. But are we truly making things better, or just bolting on features that were never needed? Niall Maher, with a wealth of experience from product leadership to CTO roles and a deep involvement in Ireland’s tech community, dives into this critical question. He challenges the pervasive notion that “AI is a strategy” and instead advocates for a fundamental shift in how we approach innovation, focusing on real value and meaningful change.

The AI Hype Train: Are We Drinking the Kool-Aid? ☕️

Niall kicks off with a playful, yet pointed, observation: “Every time you hear AI today, take a drink.” He highlights how AI has become the ultimate buzzword, often slapped onto products and strategies without a clear understanding of its purpose or impact. He questions the value being added, asking, “Are we making things better or are we just bolting [on] that never needed to be there to begin with?” This sets the stage for a deeper exploration of what it truly means to implement AI effectively.

From Startup Hustle to Enterprise Innovation: A Multifaceted Perspective 💡

Niall’s journey provides a unique lens through which to view technological adoption. His time as Head of Product in a consultancy instilled an obsession with user value, while his CTO roles in startups taught him the art of shipping products quickly. He now brings this “startup mentality” to Marsh, a large enterprise, where he leads innovation engineering.

His involvement with CoderDojo, Ireland’s largest web development community with 100,000 engineers and a million annual hits, underscores his commitment to community building and open source. This diverse background allows him to bridge the gap between rapid innovation and the structured demands of large organizations.

Marsh’s “NAI”: More Than Just an App, It’s a Platform 🏗️

At Marsh, Niall oversees their core AI platform, aptly named “NAI.” He emphasizes that NAI is a platform, not merely an application. This distinction is crucial: it’s designed to empower others to build applications on top of it, enabling the slicing and dicing of business verticals to deliver targeted value. This central platform acts as a “single pane of glass,” becoming more valuable as more functionalities are integrated.

The ROI Reality Check: Moving Beyond the Buzz 💰

While many companies are still grappling with how to demonstrate Return on Investment (ROI) from their AI initiatives, Marsh has seen tangible results. A Forbes mention highlighted an impressive 89% adoption rate for NAI, translating to approximately 100 hours saved per colleague annually. This is a significant value add, proving that when AI is implemented thoughtfully, it can deliver real business benefits.

The “Build a Hyper-Specific App” Disaster: Learning from Failure 🤦‍♂️

Niall shares a candid story of a failed attempt to build a hyper-specific app on NAI for a broker’s workflow. Despite identifying a clear opportunity for automation with an LLM, the app saw virtually no usage. The initial reaction was to blame the users, suggesting more training or mandates. However, Niall’s deeper dive revealed the real issue: a workflow problem, not an adoption problem.

The new tool, while technically faster for specific tasks, broke the users' existing workflow. They had to log into one system, get data from NAI, and then return to the original system. This added friction, making the perceived time savings negligible due to the copy-pasting involved. The key takeaway: integration into existing workflows is paramount.

Challenging the “Just Add AI” Mandate: Rethinking Strategy 🔄

Niall’s core message is a direct challenge to the prevailing “just add AI to everything” directive. He argues that true innovation requires a radical rethinking: “Instead of saying what can we automate, we have to change it up and say, what would this look like if we rebuilt it from scratch?”

He illustrates this with a hypothetical bank modernization scenario. Simply automating AI intakes (summarizing emails, parsing forms) without addressing downstream bottlenecks (like a human reviewer with an ever-growing backlog) doesn’t constitute transformation. The customer still waits, and the process remains inefficient. This is not innovation; it’s merely “cleanup.”

The Power of Disruption: Embracing Difficult Conversations 🗣️

Transformational change, Niall contends, often involves difficult conversations. At Marsh, a company with a history dating back to 1870, this means confronting legacy processes. He acknowledges that true disruption might even lead to job losses, a reality that often hinders progress.

His mentor’s advice resonates deeply: “If you haven’t annoyed somebody with your decision, you haven’t really made a decision.” This highlights the courage required to make impactful choices that challenge the status quo, moving beyond mere political maneuvering.

The Three Pillars of AI Success: Love, Memory, and People ❤️🧠👥

Niall distills his approach into three fundamental pillars:

  1. Build for Love (Consumer-Grade Experience) ❤️: Enterprise software often falls short due to a lack of user-centric design. Unlike consumer apps like Spotify, which users choose because they “feel good,” enterprise tools are often mandated. Niall argues that with AI tools for engineering, we have the opportunity to build applications that are not just functional but genuinely enjoyable to use, fostering pride in craftsmanship. He provocatively asks, “What if we charged our colleagues for it?” The implication is that if the value is undeniable, people will pay.

  2. Build to Remember (Memory Layer) 🧠: Most enterprise applications lack a robust memory system, forcing users to start from scratch each time. Consumer products, however, learn user behavior and offer personalized recommendations. Niall advocates for building memory into applications, splitting it into three layers:

    • Knowing the Material: Moving beyond simple vector search to hybrid search, blending semantic search with traditional data for accuracy, especially in regulated industries like insurance.
    • Knowing Who You Are (Personalization): Understanding the user allows for reduced context in agentic apps, leading to more efficient calls and a personalized experience. This can even lead to cutting down on tooling.
    • Knowing What You Need to Do (Data & Analytics): Collecting comprehensive data and metadata about user interactions allows for identifying bottlenecks, understanding user drop-off points, and making data-driven decisions rather than relying on guesswork.

  3. Build with People (Collaboration and Empathy) 👥: This is often the hardest part. Niall recounts a senior stakeholder’s blunt feedback: “If it doesn’t make my Tuesday easier, I don’t care what it does.” This shifted his perspective from showcasing impact to truly understanding user needs. The key is to partner with the people doing the jobs, building alongside them from the ground up. Winning over skeptics first can turn them into your loudest ambassadors. Ultimately, the goal is to make things easier and redesign workflows from scratch.

Your AI Homework: Rethink, Reimagine, and Validate 📝

Niall leaves the audience with actionable “homework”:

  • Process Reimagining: Pick one process, walk it end-to-end, and ask, “What would it look like if we built it from scratch again today?” This encourages a fundamental redesign rather than incremental fixes.
  • Value Validation: Ask yourself and others, “Would people pay for this?” This is a powerful litmus test for genuine utility, moving beyond building tools solely to protect jobs.
  • Memory Audit: If you have a memory system, audit it. Try to prove its effectiveness and identify areas for improvement.

By focusing on building AI that is loved, remembers, and is built with people, we can move beyond the hype and create truly transformative solutions.

Appendix