Related Service
AX Consulting and HR Consulting
HCG Consulting provides AX Consulting that supports HR AI transformation and HR Consulting that designs performance, organization, and culture systems.
Solve Complex HR Challenges with HCG
Talk to our experts
Resources
Before you choose a tool, draw the four-step roadmap.
"We adopted AI, but we are not sure what to do next." This is something HR team leaders say often. They have attached AI to recruiting screening and automated inquiry handling with a chatbot. What they are not sure about is what the next step is, and whether the things they are doing now point in one direction. In SHRM's survey of 1,722 HR professionals, only 46% of organizations said they would use AI in HR work in 2026, and adoption ran faster at more senior levels: 73% of those at HR director level and above had already brought AI into their work, while individual contributors stayed at around 65% (source: SHRM, 2026). Adopting AI has already become common, but few organizations have a roadmap that turns that adoption into a complete transformation.
HCG's view is that what decides an AI transformation is not the performance of the tool but a methodology that lets an organization assess its own readiness and work through defined steps in order. This guide recasts that methodology as the four steps an HR team actually walks through: assessment, pilot, operational rollout, and enterprise scale-up. It covers what to do and what to check at each step, and where each step most often stalls, in that order.
Before drawing a roadmap, the first thing to establish is where the organization stands today. Gartner's AI maturity model separates an organization's level of AI use into five stages: Awareness, Active, Operational, Systemic, and Transformational (source: Gartner). Most HR teams sit at the point of moving from the awareness stage to the active stage, or sit inside the active stage having run a few pilots. The four-step roadmap in this guide recasts HCG's five-step AX consulting methodology (strategy assessment → process design → architecture integration → operational optimization → organizational enablement) into language an HR team can use to assess and act on its own. Gartner's model and figures are referenced as evidence that this direction is not HCG's judgment alone but a pattern visible across the industry. The assessment should run along three axes.
Whether the data accumulated in HR systems is organized in a form AI can reference. If recruiting, performance, and attendance data sit in different systems and are not standardized, no amount of tooling changes the fact that the raw material for judgment is thin.
Whether there is a structure defining who reviews the judgments AI makes and who is accountable when something goes wrong. AI adopted without that structure ends up with an oversight system built only after a problem has surfaced.
Whether line managers and employees are ready to accept the change in how work is done once AI is involved. Gartner identified adjusting the operating model to fit AI as the variable with the greatest effect on AI productivity gains, putting its influence at 29% (source: Gartner). Redesigning how work is done around a tool decides the real effect more than bringing the tool in does.
Drawing a roadmap without these three assessments means the problems surface later, in the form of missing data or organizational resistance. To check the three axes concretely, work through the questions below.
| Assessment axis | Question to check | What a "no" means |
|---|---|---|
| Data readiness | Can recruiting · performance · attendance data be queried against a single standard | Standardizing the data for the work the pilot targets has to come first |
| Has the last year of data accumulated without gaps | There is not enough data for AI to learn from or reference, so the pilot scope has to narrow | |
| Governance readiness | Is someone assigned to review AI's judgments on a regular basis | The reviewer and the review cycle have to be set at the pilot design stage |
| Is there a procedure for when an AI judgment is challenged | Adopting without a procedure delays the response the first time a challenge is raised | |
| Organizational readiness | Do line managers understand why AI is being adopted | A separate communication and briefing process is needed before the pilot |
| Are there departments where resistance to changing how work is done is expected | Leave those departments out of the first pilot and address them at a later step |
If one axis produces a lot of "no" answers on this checklist, the order is to shore up that axis before starting the pilot. Governance readiness in particular is the hardest to fix later. Building an oversight system after operations have already started adds the burden of going back through judgments that have already accumulated.
Once the assessment is done, the roadmap can be designed step by step. Each step needs a clear goal, duration, core activities, and success measure, so that you can judge when to move to the next one.
| Step | Goal | Duration | Core activities | Success measure |
|---|---|---|---|---|
| 1. Assessment | Understand data · governance · organizational readiness | 2 to 4 weeks | Review data status, confirm the decision-making structure, interview the business | Readiness gaps and priority areas are documented |
| 2. Pilot | Verify the real effect in a narrow scope | 4 to 8 weeks | Select one or two repetitive, high-frequency tasks (recruiting screening, inquiry handling and the like), compare measures against the current baseline | Quantitative improvement (processing time, accuracy and so on) confirmed in the pilot area |
| 3. Operational rollout | Turn the pilot into a standing operating system | 8 to 12 weeks | Establish oversight and review, define the exception handling process, train the people involved | The pilot area runs stably without separate management |
| 4. Enterprise scale-up | Extend the proven model to other HR areas | Ongoing | Extend to similar work areas in sequence, redesign the operating model and roles | Multiple areas run at once and a regular ROI measurement cycle takes hold |
Skipping the order at any step guarantees that the problem appears at the next one. Starting a pilot without an assessment means running the pilot in an area where the data was thin to begin with, and attempting enterprise scale-up without a pilot means an unverified approach causes problems in several departments at once. Two criteria are needed together when choosing the pilot target. The more repetitive and high-frequency the work, the faster and more clearly the effect shows. Resume screening, payroll inquiry handling, and repetitive data entry fall into this category. At the same time, the area chosen should be one where failure has limited spillover. Choosing an area of hard-to-reverse decisions such as promotion or termination for the first pilot can backfire, because the cost of failure is high enough that the whole organization turns cautious about AI adoption itself. The starting point for a pilot should be an area that satisfies both criteria.
The move from pilot to operational rollout is the most common place to stall. The pilot succeeded, but the moment that success moves into standing operations, there is often no answer to the question of who will review these results every time. During the pilot the owner paid attention and checked personally, but once it becomes standing operations that attention naturally fades, and without an oversight system problems accumulate. Getting past this point requires deciding, at the pilot design stage, who will review what once this becomes standing operations.
The scale-up step brings the opposite problem, where an approach that succeeded in one area fails when applied unchanged to another. The way AI was used effectively in recruiting screening does not transfer directly to performance evaluation, because that area demands a different data structure and different criteria for judgment. At the scale-up step, every area has to go through a scaled-down version of the assessment in Section 1 again. The symptoms that appear most often at these two transitions, their root causes, and how to respond are summarized below.
| Transition point | Common symptom | Root cause | How to respond |
|---|---|---|---|
| Pilot → operational rollout | Attention fades once the pilot ends and quality gradually degrades | No one is assigned standing review responsibility | Fix the reviewer · review cycle for the operational stage during pilot design |
| Pilot → operational rollout | No standard for handling exceptions, so each owner responds differently | The exception handling process was not defined in advance | Write a handling manual based on the exceptions that came up during the pilot |
| Operational rollout → enterprise scale-up | The approach that worked in one area is applied unchanged elsewhere and produces no effect | Overlooking that data structures · judgment criteria differ by area | Run a scaled-down Section 1 assessment for each area being added |
| Operational rollout → enterprise scale-up | Scale-up accelerates and the governance built early is overloaded | Governance was designed around a single area without considering expansion | Design governance from the start on shared criteria that apply across multiple areas |
The pattern repeating in this table is that the problems at each transition mostly start from a design that did not account for the next step. Keeping operational rollout in mind while designing the pilot, and enterprise scale-up in mind while designing the operational rollout, avoids a good share of these problems in advance.
Designing and running these four steps from scratch with an HR team's internal resources alone is difficult in practice, because the methodology the assessment requires, the experience of designing a pilot, and the governance design at the point of operational rollout each demand different expertise. Consulting's AX consulting designs and runs all four steps as a single flow. The assessment in Section 1 begins in the strategy assessment (Strategy) stage of AX consulting, measuring data · governance · organizational readiness with a structured methodology. If the assessment shows that data in a given area is not standardized, AX consulting designs that standardization directly in the process design (Design) stage. The pilot in Section 2 and the governance at the transition points covered in Section 3 are handled in the architecture integration (Build) and operational optimization (Run) stages, and the standards and criteria designed there are implemented in systems through the data structures of hunel · JaDE · talenx or the Agent capabilities of elizax. Because this runs inside a single consulting flow that continues through organizational enablement (Enablement), rather than assessment, system build, and adoption proceeding separately, less information is lost each time responsibility changes hands between steps. The first thing an HR team preparing for AI transformation should do is not choose a tool but draw the roadmap that walks through these four steps in order. Without that roadmap, the gap between organizations that stopped at the pilot and organizations that completed the transformation only widens with time.