# How to Evangelize AI Adoption in Enterprise Organizations By Aby Varma · Published 2026-04-18 · Updated 2026-11-02 · Professional AV on MarketScale Canonical: https://www.marketscale.com/industries/pro-av/how-to-evangelize-ai-adoption-in-enterprise-organizations Creator hub: The Marketing AI SparkCast > Enterprise AI adoption is an enablement problem, not a technology problem. David Kuoch, VP and AI Evangelist at Citi, joins Aby Varma on The Marketing AI SparkCast to explain how live demos, prompt engineering frameworks and scalable operating models drive adoption across a large organization. ## Key points - Enterprise AI adoption is fundamentally an enablement problem, not a technology problem - Kuoch emphasizes live demos and prompt frameworks to help employees use AI effectively. - Shared operating models can improve consistency and business value as AI use scales. Most enterprises have already handed employees AI tools. Far fewer have gotten those employees to use them well. On this episode of The Marketing AI SparkCast, host Aby Varma, Founder of Spark Novus, talks with David Kuoch, Vice President and AI Evangelist at Citi, about why enterprise AI adoption is an enablement problem, not a technology problem. Kuoch works at an unusual intersection. His background is in project and program management, he works with Citi's engineering team to solve complex problems with AI, and he also sits on the data governance side. As an AI evangelist, his job is teaching the rest of the organization "how to fish, so to speak." Scale shapes everything he describes in the episode: about 220,000 people at Citi, roughly 15,000 of them in enterprise data operations. In the last year he has presented to about 5,000 colleagues and run more than 50 demos. Smaller companies can move fast and break things. A financial institution has to live within its governance framework. "We, unfortunately, can't break anything," Kuoch said. ## Raise the floor, then raise the ceiling Kuoch frames evangelism around two goals that coexist. Raising the floor means bringing people at every skill level to the point where they understand what AI can do. Raising the ceiling means delivering business value: moving from proofs of concept to something that scales over the long term. Access alone doesn't produce adoption. Giving people tools "doesn't necessarily mean it's gonna be adopted," he said, in part because of an inherent fear that AI will replace people. His answer to that fear is blunt. > AI is gonna replace the people that don't know how to use AI. — David Kuoch, Vice President and AI Evangelist, Citi The practical starting point is structure. Look at your day-to-day work, find the manual, time-consuming, low-hanging fruit, and start there, whether you sit in marketing, compliance or finance. ## Show the work, not the slides Education comes first among the 10 principles in Kuoch's AI enablement approach, and his method is the live demo. He takes an example of what someone actually does in their workday and does it with them on the spot. > I'm not a big fan of PowerPoint slides... What I like to do is when I sit on these demos, I'm gonna show them exactly the work that needs to be done. — David Kuoch It resonates more, he explained, when people apply AI to their own work instead of someone else's. From there, his framework covers problem framing (start with the desired outcome, whether that's a report or a metric), prompt engineering (he teaches the CO-STAR structure of context, role, objective, style, tone and response, because junk in means junk out), process decomposition (turning something messy into something structured that can scale), data interpretation (using AI to turn static data into dynamic data), and finding the most time-consuming steps in a workflow where AI belongs. Governance runs through all of it. Kuoch is emphatic that AI is advisory. > It's meant to be your assistant, your copilot to sort of get you to do 60 to 80% of the work. — David Kuoch The rest belongs to the subject matter expert who validates the output. Inside Citi, he said, the work is never fully automated. There are checks and balances so the organization doesn't rely solely on AI to make decisions. ## From prompt libraries to operating models Varma raised the tension every enterprise hits at scale: standardization versus the flexibility people need to use AI in the context of their own jobs. Kuoch said the first thing that appears organically is a prompt library, which is a good start. The trouble comes when thousands of people build their own libraries in silos and end up recreating the wheel. That phase feels like success because people are using the tools, but consistency suffers. Once an organization matures, he said, it has to step back. > We need to shift the mindset and think about playbooks, intelligence level, operating models. — David Kuoch With a shared operating model, anyone running the same playbook can produce the same report. That is where consistency and business value show up. Instead of ten versions, teams agree on one model and tweak it for their needs. Kuoch designs for the widest funnel first, then lets groups such as compliance and risk adjust tone, policies and procedures in modular layers, with version control so each iteration stays familiar. Varma said he sees the same shift with CMOs and marketing clients moving from wild-west experimentation into production, and warned against rigid rollouts. > It's a square peg round hole type situation. You force teams in it. Those things are never sustainable. — Aby Varma, Founder, Spark Novus ## Don't skip the growing pains Asked whether organizations should skip early experimentation and jump straight to mature use cases, Kuoch pushed back. He went through it himself; by his own count, it took about 40 tries to get something right. Skipping that phase leaves a gap where people don't understand how the technology works. > I think it's healthy for organization to struggle and get people to do that. — David Kuoch For business value, his advice is to focus on one desired outcome and work backwards, almost like reverse engineering. When information comes in for the first time, build structured data from it. That core data can then become a PowerPoint slide, an executive summary or whatever comes next, without reprocessing. Community keeps the learning moving. Kuoch is part of an internal group of about 4,000 people that shares what's new and which techniques work, then passes it on to colleagues who are new to AI. Repetition is part of the job. In an organization that size, someone will hear about a six-month-old idea for the first time, so patience matters. ## What smaller teams can take from it Smaller organizations have a luxury enterprises don't: they can iterate and move fast. "It's easier to teach 30 people AI skills than 200,000 people AI skills," Kuoch said. Even so, he advised designing guardrails and governance early, so AI isn't left to make decisions on its own. He also urged openness: run demos for your team, create a chat where people share what they learn, and share your designs so someone else can build something better. > AI is not a secret. Share what you actually have. — David Kuoch Looking ahead, Kuoch believes AI is still in its infancy and that anyone listening is an early adopter. His closing advice is to look at your own workflows first and start with the low-hanging fruit. "It's less about the tool, and it's more about the system integration," he said. Watch the full conversation between Aby Varma and David Kuoch in the episode video, or listen to The Marketing AI SparkCast on Spotify. ## Sources - [How to Evangelize AI Adoption in Enterprise Organizations (Spotify)](https://open.spotify.com/episode/5zNkNBXU8NTSK1Axm9Cwvd) (The Marketing AI SparkCast) - [How to Evangelize AI Adoption in Enterprise Organizations (episode video)](https://marketscale-4.wistia.com/medias/hiz9ursfjx) (The Marketing AI SparkCast) Tags: Aby Varma, The Marketing AI SparkCast, Spark Novus, David Kuoch, Citi, AI Adoption --- Source: MarketScale, https://www.marketscale.com/industries/pro-av/how-to-evangelize-ai-adoption-in-enterprise-organizations. Published for AI indexing and citation; cite the canonical URL. Site guide for agents: https://www.marketscale.com/llms.txt