How We Built an AI-Ready Organization: Inside One Company's First Year
Any opinions and errors my own, not the company’s.
It’s a wild time to be a business leader. The market is saturated with AI transformation claims and FOMO-induced decisions. When we started, there were plenty of headlines about what not to do but no blueprint for what to do. One year in, there is much still left to be done — but we’ve made quiet and genuinely exciting progress.
Here’s what we built at Omnidian and what we learned.
Context
Every company claims to care about culture, but at Omnidian it is the beating heart of the organization. The founders and many of the employees are here out of a genuine commitment to the mission and fighting climate change. It is a scale-up outgrowing scrappy start-up origins, navigating the question of how to utilize technology while retaining the human touch.
From the outset I knew I wanted to avoid corporate initiative failure modes. AI transformation at Omnidian would be people-centered from day one, vision-led rather than exclusively force-fed or bottoms-up, with a visible executive champion from the start. I put my hand up for that role.
Starting with an honest account of where we were at the beginning matters. We simply had no way to know who was using what AI, when, and how. So we ran a survey in May 2025 in which over 90% of the company participated. The findings told us we had a lot of work to do — and directly informed what came next:
80% felt at least “somewhat optimistic” about the impact of AI on their work.
Sub-pockets of the company had entirely different AI cultures — avid experimentation in some teams, distrust and non-use in others.
Employees had poor visibility of what their colleagues were doing, with people noting “I bet everyone uses AI” and “I don’t think many people use AI” within the same team.
Almost 40% reported using AI “rarely” or “never.”
Only 11% rated our AI utilization as effective.
Employees asked for hands-on practical training, policy clarity on what was safe, and they expressed genuine excitement about innovation alongside real concern about job security and human oversight.
That survey did not just tell us where we were. It became the foundation of our roadmap.
Legal as AI Champion: The Unexpected Benefits
One of the things I am most grateful to our CEO for was his full-throated support for having a lawyer serve as executive champion for AI. We did not have a dedicated budget or headcount for this initiative, but a clear mandate from the top enabled everything that followed.
Many team members started off skeptical that legal and innovation even belonged in the same sentence. They came around quickly as we discovered the following:
Credibility: Legal has an ethical duty to the organization that transcends tribalism especially on sensitive issues. In many organizations, AI has become controversial. Legal can come with solutions that are mission-first and risk-calibrated to the business. When Legal says “we have thought through this and it is workable,” it lands differently.
Connection: Legal is the connective tissue across the organization. It is one of the few functions that interfaces with every other team — which means a unique understanding of use cases, risk, and goals that another function might miss.
Velocity: Governance can finally move at the pace of the business. When Google introduced agentic AI capabilities into Gemini, our AI policy was updated the same day. Governance reacting to last quarter’s updates will lose trust and get ignored.
Legal-as-dinosaur rep: When legal is telling you to speed up, maybe you’re going really slow!
The most practical benefit was clarity. The most frequent adoption blocker we heard in the May survey was “I don’t know what’s safe, and that is preventing me from moving faster.” When Legal owns governance and is paying close attention, that question gets answered fast.
Our Approach: People First
We put particular emphasis on people and governance — the two dimensions most programs ignore. They are also causally linked: you cannot build genuine AI adoption without first understanding where your people are, and you cannot move fast without governance that answers the questions that are slowing them down. We worked both in parallel from day one.
GETTING YOUR PEOPLE AI-READY
Start with feelings, not tools
The May survey did more than give us data. It signaled to the organization that their experience of this transition mattered — that we were not simply mandating adoption and measuring logins. We held unscripted Q&A sessions across the company, asked for the good, the bad, and the ugly, and gave honest answers. When we listened, people gave really thoughtful input.
Survey findings shaped the roadmap directly. This is not standard practice in AI adoption programs, which tend to be designed top-down and then pushed into the organization. We listened first, then built.
AI as career boost, not job displacement
The framing question matters more than most leaders realize. “AI will make your job easier” lands very differently from “AI will replace jobs” — but employees have been reading the headlines and they have formed views. We were deliberate about the message: AI is here to make space for your highest-judgment work. The tasks that drain your energy without requiring your expertise are the ones we want to automate. This then makes space for the work that only you can do.
We connected incentives explicitly: build AI into your workflow, grow your capability, grow the business. This framing also triggered important internal questions we had to answer honestly — how centralized did we want to be as an organization, and how did we want to recognize and reward innovation coming from the ground up? We grappled with these questions head-on as an executive team, informed by team needs.
Training: both dimensions matter
Most AI training programs do one of two things: they teach people how to use the tools, or they cover responsible AI and governance. We did both, and we believe both are necessary. Skills without guardrails produce ungoverned risk. Guardrails without skills produce compliance theater. The combination — here is how to use these tools, here is how to use them responsibly — is what actually moves people from reluctant compliance to genuine adoption.
INCUBATING INNOVATION
Building a culture of sharing, not siloes
Innovation stalls when people hoard what they discover. We built in structural incentives for sharing: regular show-and-tells where employees demonstrated what they had built or figured out, Slack channels dedicated to AI use cases and experiments, and a genuine norm of celebrating attempts — including the ones that did not work. Psychological safety around experimentation is not just a soft benefit, but the precondition behind the flywheel of innovation, which is often iterative.
The Innovation Task Force — and why timing is everything
Early on, we established an AI Innovation Task Force — a cross-functional group of volunteers charged with bringing simple AI solutions to productivity blockers across the organization. Requesters needed to only bring ideas and the task force would do the rest (design, test, launch). The intentions were good, but the results were limited. We were asking employees to innovate with a technology most of them had never actually used. It is a bit like asking someone to brainstorm new recipes before they have ever cooked.
The task force produced isolated unlocks but we needed a breakthrough: widespread, genuine engagement with AI tools at the individual level. That required a different intervention.
The hackathon as the unlock
The hackathons – run first with the executive team, then with 150 people at the company onsite – changed things. Bringing people together in-person to actually use AI tools in pursuit of real problems they cared about produced something the task force could not. People who had never touched an AI tool came out with working prototypes. People who thought AI was not relevant to their function found applications they had not imagined. Teams competed for prizes, and we could not get teams off to their next session because they were glued to their seats trying to build. The energy was different, and outlasted the main event which was just 45 minutes.
The results were shared across subsequent team meetings, show-and-tells, and grew organically without intervention from a task force. The sequencing lesson is clear: hands-on experience before ideation, not the other way around. You cannot think of ideas for a technology you have not touched, and it starts from the top.
Governance: The Enabler, Not the Brake
Lead with governance — it accelerates adoption
This is the counterintuitive finding: governance, done well, does not slow adoption down. It speeds it up. The most common adoption blocker we heard in the May survey was “I don’t know what’s safe, and that uncertainty is stopping me from trying.”
So governance came first. We drafted an AI policy written specifically to answer that question: not to cover legal bases, but to give employees a clear answer to “can I use this?” Everyone signed it. We communicated it actively, not as a one-time rollout but as a living document. And we iterated it publicly. When the landscape changed, the policy changed with it, visibly and quickly.
AI Governance Committee and board reporting
Responding to May survey input asking for a cross-functional leadership group thinking deeply about AI, we launched an AI Innovation and Oversight Committee. This was not a compliance checkbox. It was the mechanism through which AI decisions got made with appropriate visibility and accountability. The committee met at first regularly then on an as-needed cadence to discuss first-of-kind issues on AI innovation and oversight – a dual mandate that was baked in from the start. The goal was to ensure that governance was visible at the highest level of the organization and not siloed inside a single function.
Procurement framework
The procurement question was equally urgent: we had no visibility into what tools were already in use across the organization, or whether they were safe. Latitude without guardrails and patchwork experimentation made things individually easy, but collectively was reducing trust. We built a lightweight procurement framework designed to answer that — clear enough to give people guidance, fast enough not to become a bottleneck. The goal was not to restrict what people could use. It was to make sure we knew what was in the stack and had made a considered decision about it.
Knowing when to say no
Clear nos are as important as clear yeses. We said no to third-party notetakers — the privacy and data security implications were not ones we were willing to accept without a more thorough assessment. We drew a line on personal AI subscriptions being used for work purposes without approval. We said no to certain vendors not aligned with our values. These decisions were not popular with everyone. But they were consistent with our stated values and they reinforced that the governance framework was real, not performative.
What the Foundation Unlocked
There is a reason people and governance dominate year one. They are the foundation that makes everything else possible. Once employees trust the direction, understand the guardrails, and have genuinely used the tools — the question changes. It stops being “is this safe?” and “should we be doing this?” and becomes “what can we now build that we couldn’t before?” Process redesign and technology investment sprint faster on a foundation of cultural readiness and governance clarity. The organizations that try to skip to the sprint first spend most of their time managing fear and cleaning up ungoverned risk. We did not skip. Here is what became possible once we had the foundation in place.
PROCESS
Small wins before big swings
We deliberately sequenced early AI initiatives around achievable wins rather than transformation projects. This was a strategic choice, not a conservative one. Early wins build the organizational muscle memory for using AI, they create visible proof that this is real and useful, and they give people the confidence to try harder things. Going straight to big swings — before trust is built and capability is established — is how you lose the organization before you have started.
One specific example was the time the team spent waiting on review requests for third-party NDAs. This was a low-judgment high-attention workstream for legal and as such was a perfect candidate to build an AI agent. After a careful pilot period it has gone live across sales and other business teams, bringing average review time down from weeks to seconds. I retain human oversight by building in escalation paths directly into the agent output, enabling unblocks after a quick glance.
Who decides what
We drew a deliberate line between two types of AI decisions. Individual employees own the question of how to sharpen their own work — finding AI tools that help them do their specific job better, experimenting with workflows, building personal capability. Product and engineering own the big technological swings: platform decisions, architectural choices, integrations that affect multiple teams.
This split reduced friction significantly. Without it, every AI decision became a negotiation about authority and approval. With it, people knew what was sanctioned for individual experimentation and what required a different conversation. Clarity on accountability, it turns out, is itself an adoption accelerant.
TECHNOLOGY
Official tool rollout
In May we had no official AI tool, so enabling one with guardrails was a critical first priority to eliminate usage of at-risk activity on personal subscriptions. We aligned early on a unified foundation and modular architecture layered on top based on team needs. Gemini was made available shortly afterwards, and the team was empowered to build a tech stack tailored to their needs.
Rolling out Gemini did not magically make people use it. We pay attention not only to adoption metrics but create scaffolding to ensure the adoption is meaningful and moves outcomes. The gap between a tool being available and a tool being used is not a technology problem. It is a people and communication problem. How you introduce a tool — what problem it solves, why it was chosen, what it does not do — shapes whether people engage with it genuinely or treat it as another mandate to manage around.
Thinking about the AI stack like a portfolio
Every AI tool purchase in this environment feels a little like an investment decision. The underlying technology is moving fast, the dominant platforms are not yet determined, and some of what exists today will not survive. We talked to our teams explicitly about this: not to create anxiety, but to build the kind of informed judgment that prevents lock-in.
The questions we kept coming back to: Are we buying into a vendor whose position in the stack could get displaced? Are we building dependencies that will be expensive to unwind? How does this tool interface with the others we are already using — and increasingly, how does it behave alongside other agents in our environment? That last question matters more than most procurement frameworks account for. As agentic AI becomes more common, the interactions between tools are not just a compatibility question. They are a governance question.
Our governing principle was straightforward: where our competitive advantage lives, we build and protect it. Where it does not, we buy the best available tool, move fast, and stay light on our feet. That framework did not eliminate uncertainty — nothing does in this market — but it gave teams a basis for making decisions without escalating everything.
Where We Are Now + Why This Matters For Everyone
One year did not solve AI for the entire org. Far from it. But we’ve made strides that seemed impossible twelve months ago:
85% employees active users of AI with 65% using at least weekly
60% rating our AI utilization as effective
5+ hackathons
20+ agents in use
NDA review time reduced by 99%. A 150-person agentic AI workshop at our company onsite where beginners and power users alike built working agents. An organization that talks about AI differently than it did a year ago — with less fear, more curiosity, and a clearer sense of what it is for.
We will probably never call AI transformation done, but have done the work of opening hearts and aligning minds on some fundamental questions in the first year. Conversations are still ongoing, programs are still being designed, and there are still teams that are much further along than others. This is not a caveat. It is the point.
The patterns here are not company-specific. They are structural. The organizations getting AI adoption right share the same characteristics: people-first, vision-led, governance-enabled. The ones struggling share the same failure modes: tool-first, bottoms-up strategy, governance as an afterthought.
If you are building this inside your organization, the priorities are the same regardless of your industry or size: listen before you mandate, lead with governance because it removes blockers rather than creating them, and sequence for trust before you sequence for transformation. The technology is available to everyone. The organizational capability to use it well is not. That is the durable competitive advantage. The companies that figure it out in the next two years will have built something genuinely difficult to replicate — not because of the tools they chose, but because of the culture they built around them.
Natalie is VP Legal at Omnidian, where she led a full-arc enterprise AI adoption.
The Judgment Layer publishes on AI governance, board accountability, legal intelligence, and the judgment no algorithm is taking from you.