Building an AI-native workforce inside an approximately $3 billion technology company.
We are leading an AI transformation program for a global technology company valued at approximately $3 billion. The goal is not simply to give employees access to another AI tool. The goal is to change how teams work, help them identify where AI can take on meaningful responsibility, and give them the skills to build secure systems around real business workflows.
The engagement combines company-wide education, hands-on working sessions, technical discovery, and team-specific agent development. Employees learn the foundations first, then apply them directly to work they already own. This keeps the program grounded in practical outcomes rather than generic demonstrations.
The current work includes an HR agent in development. That build sits within the wider program: understanding the team's process, working through security constraints, and keeping human responsibility clear. It is work in progress, not a claim that an autonomous HR function is already deployed.
| Project context | Public detail |
|---|---|
| Client | Global technology company · name withheld |
| Approximate valuation | $3 billion |
| Scope | Company-wide foundations and team-specific development |
| Current build | HR agent · in development |
| Program status | In progress |
The company already had technically capable teams and access to advanced AI products. The problem was uneven adoption. Some employees were experimenting regularly, while others still treated AI as a chat interface for isolated questions. Teams did not share a consistent understanding of what agents were, where they could be trusted, or how to connect them safely to internal systems.
There was also a gap between individual productivity and organizational change. Helping one employee write faster is useful, but it does not automatically improve how a department operates. Lasting value requires teams to redesign workflows, clarify approvals, document edge cases, and decide where human judgment must remain in the loop.
Security added another layer. The company handles sensitive information and operates within a complex technical environment. Any useful implementation had to account for permissions, personally identifiable information, data boundaries, existing infrastructure, and internal review processes. A public workshop or off-the-shelf training course would not have been enough.
The first stage is a four-part foundation series. These sessions establish a shared vocabulary and move participants beyond basic prompting. Employees learn how to work effectively with modern AI tools, organize reusable context, create skills and projects, use coding agents even when they are not software engineers, and understand how autonomous agents differ from standard chat systems.
Participants do not just watch a presentation. They complete exercises, test tools against their own responsibilities, and leave with homework tied to work they already perform. Sessions are recorded, but live participation is encouraged because the discussion often surfaces concerns and opportunities that would otherwise remain hidden.
The second stage focuses on individual teams. We ask the people closest to each workflow what they want to build, how the process works today, what information moves through it, where it breaks, and which decisions require human judgment. A workflow that looks repetitive from the outside may contain dozens of exceptions known only to those people. Rather than automate an incomplete diagram, we capture those exceptions and design around them.
The third stage is hands-on development. Teams work through selected use cases during longer training and working sessions. Depending on the workflow, the output may be a reusable AI skill, a personal agent, a structured process, or a prototype connected to an approved internal system. The objective is for each group to leave with something useful, not merely a list of ideas.
During the foundation phase, participants attend a one-hour instructional session early in the week and a 90-minute working session later in the week. Between sessions, they complete focused assignments such as testing a workflow, creating a skill, reviewing a resource, or documenting a process.
During the cohort phase, each team begins with a discovery session and then joins a 90-minute build session. Short individual or small-group calls are added only when a workflow needs deeper technical or operational attention.
Dedicated communication channels allow stakeholders to answer follow-up questions asynchronously. This reduces unnecessary meetings while giving our team access to the people who understand each process. It also creates a written record of decisions, constraints, and unresolved questions.
This account describes the current engagement and program design. The work is in progress; the milestones below are success criteria, not a list of completed outcomes.
Security is built into the engagement from the beginning. Before teams connect agents to sensitive systems, we work with technical and security stakeholders to define what data can be used, which integrations are permitted, and where additional review is required.
We teach participants to think in terms of scoped permissions rather than unlimited access. Agents should receive only the tools and information needed for a specific responsibility. High-impact actions should require human approval, and important activity should be visible enough to review later.
This approach helps the company move quickly without treating experimentation as an excuse to bypass governance. It also gives employees a clearer understanding of why a prototype that works in a personal account may need a different architecture before it can be deployed across the business.
Employees need permission to rethink work, time to experiment, and confidence that leadership understands the difference between a failed test and a failed strategy. Managers need a realistic picture of what AI can handle today, where it remains unreliable, and how responsibilities may change as systems improve.
For that reason, the program involves leadership as well as practitioners. Leadership alignment can happen through live sessions or concise asynchronous communication, but the message must be consistent: adoption succeeds when teams are encouraged to change how work gets done, not when a new tool is added to the existing process and ignored.
This engagement is still in progress, so we are not presenting projected benefits as completed results. The initial milestones are concrete: establish a shared AI baseline, identify viable workflows, document security constraints, create working prototypes, and develop internal champions who can continue the work.
Longer term, success means that teams can recognize good agent opportunities on their own. They can separate impressive demos from dependable systems, estimate the operational value of a workflow, and design human approval into the right places. The company gains more than a collection of automations. It develops the internal capability to adopt AI repeatedly and responsibly.
For an organization of this scale, that capability matters. Even modest improvements in how thousands of decisions, handoffs, documents, and recurring tasks are handled can create significant capacity. That is the opportunity this program is working toward, not a measured result we are claiming today.
The larger opportunity is not one agent or one productivity metric. It is a workforce that understands how to delegate work to AI while keeping people accountable for the outcome.