A 12-month plan to reskill your whole company to work alongside AI

The short version

When training isn't tied to real work, everyone is back to working the old way within two weeks of it ending.
Start with the work that is really changing and design the learning around it, instead of starting from a course calendar.
What proves people have improved is real work that can be measured before and after, rather than a certificate.

Month 0 to Q1: build the foundation from strategy and work

Sending the whole company to the same one-day training is the easiest approach to organize and the least effective. People who come back with no work to apply it to right away forget it within two weeks. The right starting point is to look first at which work is about to change, and then teach the people who do that work.

For the development team · Technical detail

Start with business goals, such as increasing capacity, cutting lead time, reducing waste or opening new revenue. Then build a task inventory that sorts work into Augment, Automate, Human-only and Stop. Create workforce scenarios showing which work will grow, which will shrink and which skills are missing. Don't set a target of “100% trained” without a destination workflow.

A team sorting the work of one position into Augment, Automate, Human-only and Stop groups — Start from a task inventory of real work and leave the training calendar for later. Once you know which work will change, you know which skills to teach.

For the development team · Technical metrics

In Q1, everyone completes AI literacy, verification, data safety and the incident reporting channel, using the company's own problems. Open a tool catalog and office hours to reduce shadow AI. Record a skill baseline and the current work metrics. Choose 3 to 5 pilots, each with a sponsor, a process owner, HR, IT/security and employee representatives.

Design principle: Learn during working hours, practice on real work, and have managers learn alongside their teams, because managers are the ones who change workflows and KPIs.

Q2: Role Academy and capstone

To prove who can really do the work, use the same principle as hiring: look at the work itself. Have each person take a piece of their own real work, redo it, and measure how much the before and after differ.

Role	Must be able to	Passing piece of work
AI User	Brief, verify, protect data	Real work that passes the checklist
Practitioner	Templates and workflows	Less time with the same quality
Reviewer	Check risks and exceptions	A test set and escalation criteria
Builder	Connect systems, evaluate, monitor	A pilot with logs and a fallback
Owner	KPIs, SOPs, budget, incidents	A business case that can be signed off

Learners complete a capstone based on their own work, compare before and after, and hand over a template, checklist or SOP for others to use. Certification comes from results; attendance alone doesn't earn it. Build champions in each department to adapt the central standards to their own context.

An employee presenting a capstone based on real work, with before-and-after comparisons — A certificate is not evidence. What proves a skill is real work with measurable before-and-after results.

Don't add to the workload: If people learn AI but still have to produce all the old reports and follow all the old methods, retire the old steps once the pilot passes. Otherwise adoption turns into overtime.

Q3 to Q4: change the people systems and scale what works

For the development team · Technical metrics

In Q3, update job descriptions, KPIs and career paths. Add roles such as reviewer, process owner, knowledge curator and automation champion. Open an internal marketplace so people who have passed their capstone can join cross-department projects. Set aside official time for improvement work instead of pushing all the learning to after hours.

Employees choosing new roles from an internal opportunities board, with their managers weighing in — When job descriptions, KPIs and career paths change with it, learning becomes a career path rather than a one-off course.

For the development team · Technical detail

In Q4, scale use cases with a single playbook: process, data, risk, evaluation, training, support and value measurement. Retire duplicate tools and projects that don't pay off. Redo the workforce scenarios with real data, and plan redeployment, reskilling, natural attrition or hiring according to the gaps, treating people fairly and in line with the law.

What needs to be ready before you scale

The workflow consistently meets quality standards and has an owner
The roles of people and AI, and the escalation paths, are clear
Support, monitoring and a running budget are in place
Jobs and KPIs don't conflict with the new way of working
There is a path for people to move into new roles and build their skills

Transparency, governance and a scorecard

Communicate what the company is testing, what usage data it collects, who can access it and whether it is used to evaluate individuals. Provide a way to appeal when AI affects people. Let teams share failures without being punished, and don't make promises about headcount that the organization can't control. Being straightforward builds more trust than saying “nothing will change.”

Dimension	Metric
Capability	People who have passed real work for their role
Adoption	Workflows in continuous use
Business	Cost, cycle time, quality, capacity
Talent	Internal mobility, time-to-proficiency
Trust & Risk	Incidents, overrides, confidence

Review each quarter what should be scaled, improved or stopped, and update the curriculum based on incidents and new workflows. Reskilling doesn't end in month 12, but the first year should build a system that can learn from real work on its own, with an owner, a community, a library of standards and an ongoing budget for improvement.

CHANGE NOTE · RESISTANCE ISN'T ALWAYS ABOUT AI

Sometimes employees are resisting the duplicate work added in AI's name

If people have to study after hours, use the new system during the day and still keep the old Excel file just to be safe, choosing not to adopt is understandable. A reskilling plan needs a decommission plan that states when the old method will stop, who will help during the transition and which KPIs will be adjusted.

Hypothetical case: An accounting team trained to read invoices with AI, but at first they had to check every field, so it felt slower than before. The project changed the system to highlight only values with low confidence, and counted checking time as part of the learning KPI. After six weeks, the team began to see less repetitive work. Their confidence came from good design rather than from speeches.

The 60-30-10 formula

60% of learning comes from real-work capstones, 30% from coaching and community, and 10% from foundational content. Courses provide the vocabulary and the safety rules, but fluency comes when employees solve their own problems and get feedback.

Tip: Every quarter, ask two separate questions: “Do you use the tool?” and “Which part of your work has actually improved?” The first answer measures adoption and the second measures value. Don't mix the two.

DNA MAKER · SOLUTION BLUEPRINT

From knowledge to a system that solves the problem in practice

The core of the problem

Reskilling won't succeed if the old jobs, KPIs and workflows all stay in place. Learning has to lead people into new roles and new ways of working.

A step-by-step approach

Connect business strategy with the task inventory and workforce scenarios
Build a 60-30-10 Role Academy with capstones and manager coaching
Adjust jobs, KPIs and career paths, then scale only the workflows that deliver business results

Build the digital structure that carries learning back into work and careers

Reskilling shouldn't be a training project cut off from how work actually gets done. DNA Maker works with executives, HR and team leads to connect the task inventory, roles, capstones and feedback, so it's clear which work people are learning to do better. Knowledge of talent and career paths stays with the organization. Our part is to design the learning journey and the workflows that show employees what to learn, where to practice, who reviews their work and how work that passes gets put to use.

The solution might be a Learning & Workflow Portal, an AI practice sandbox, a knowledge hub or a dashboard for capstones and adoption, connected to the tools the organization has approved. DNA Maker can help with everything from experience design, prototyping, web and mobile development, AI agents, integration and analytics to using AI Autonomous Development to speed up delivery under proper review. If you have a people development roadmap but can't yet picture the system that supports it, we're ready to talk through a structure that starts small with one role and expands once the results are proven.

SOFTWARE ENGINEERING GLOSSARY

Software engineering glossary

This table isn't meant to be memorized. It helps executives, process owners and the development team talk to each other without reading the same words differently. Read the meaning, the example and the question on the right, because these questions often bring hidden scope, risks and costs to light before development starts.

Term	What it is	A simple example	What to ask the development team
Learning Platform	A system that organizes lessons and tracks development	Employees see a learning path for their role	How does the system connect learning to real work?
Sandbox	A practice area kept separate from live systems	Practicing with agents without touching production data	How is test data kept separate from production, and how is it cleared?
Capstone	A real project used to prove a skill	Cutting report preparation time while still passing the quality check	Which piece of work proves the skill can really be applied?
Adoption Analytics	Data on how people use a system	Looking at which workflows get reused, beyond simple logins	Are we measuring usage, or the value that comes from usage?
Decommission	Retiring an old system or method according to a plan	Shutting down the old Excel file once the new workflow is stable	When will the old method be shut down, and is there a rollback plan?

Try this tomorrow: Pick one department, build its task inventory and define one capstone with a baseline. Starting from the work makes the 12-month plan concrete right away.