The CLO's Guide to AI Upskilling in 2026: What to Buy, Build, or Skip

The landscape every CLO is navigating right now

Chief Learning Officers in 2026 are operating in one of the most demanding skill development environments in the history of the profession. The pace of change in AI tooling is faster than any previous technology wave. The skill gaps are real, urgent, and distributed across every function. The pressure from the C-suite to "do something about AI capability" arrived before most L&D teams had the budget, vendor landscape knowledge, or internal expertise to respond intelligently.

The result is a wave of AI training investment that is producing highly variable results. Some organisations are compounding genuine capability. Others are funding completion-rate theatre — spending meaningful budget on programs that look good on dashboards but change nothing about how work gets done.

This guide is for CLOs who want to be in the first category. It covers the buy vs. build vs. skip decision framework for AI training investment, the functions where AI upskilling is most urgently valuable, what to look for in an AI training provider, and how to build the internal business case to secure and maintain the budget.

The landscape in numbers: what you are actually dealing with

Before making investment decisions, it helps to have a clear-eyed view of the environment:

• Tool proliferation is real and accelerating: The number of AI tools relevant to business functions has roughly tripled since 2023. Most employees are simultaneously overwhelmed by choice and under-supported in making those choices. L&D's role is to cut through the noise, not add to it.
• Skill gaps are function-specific: Engineering teams need a fundamentally different AI curriculum than marketing teams, which need a different curriculum than operations. Generic "AI literacy" programs that treat all functions the same produce weak outcomes across all of them.
• The half-life of AI training content is short: A curriculum built around specific tool features from 18 months ago is already partially obsolete. CLOs need training partners who update content continuously, not annually.
• Internal champions exist but are underutilised: In almost every organisation, there are 3–5 employees who have been deeply experimenting with AI tools on their own time. These people are an enormous asset that most L&D programs ignore entirely.

The buy vs. build vs. skip decision framework

Not every AI training need should be addressed the same way. Here is the framework we recommend for making the buy/build/skip decision:

Buy: when the need is urgent, specialised, and external expertise is ahead of yours

Buy external training when:

• The function requires deep, current knowledge of a rapidly evolving tool ecosystem (engineering AI tooling, AI agent development)
• The training needs to deliver measurable outcomes on a defined timeline with accountability
• You need practitioner instructors who are actively building with these tools — not just certified to teach them
• You want an external benchmark for what "good" looks like, rather than building capability in a vacuum

Engineering AI upskilling (Claude Code, Cursor, AI agents), product AI training, and advanced GTM AI workflow programs almost always fall in the "buy" category for most organisations. The tool ecosystem changes fast enough that maintaining internal curriculum is genuinely expensive.

Build: when the need is ongoing, foundational, and deeply context-specific

Build internal training when:

• The capability is foundational and stable enough that a curriculum written today will still be relevant in 12 months (e.g., AI prompt principles, responsible AI use policy)
• The training is deeply specific to your internal systems, data, and processes in ways that an external vendor cannot replicate
• You have internal champions who can credibly deliver the training and who are committed to keeping it current

Good candidates for internal build: AI use policy training, internal tool onboarding (your custom AI integrations), and AI champion communities of practice.

Skip: what to stop wasting budget on

The AI training market is full of programs that look credible but deliver minimal impact. Skip:

• Generic "AI literacy" e-learning modules that could have been written two years ago and probably were
• AI awareness programs that end before participants have done anything hands-on with a real tool
• Vendor-funded training that is primarily a product demo disguised as education
• Any program whose "outcomes" are defined entirely in terms of completions, certificates, or satisfaction scores

Which functions need AI training most urgently — and why

Budget is finite. Prioritise investment by expected impact:

1. Engineering (highest urgency)

Engineering teams have the highest leverage for AI upskilling because the tools are most mature, the productivity gains are most measurable, and the compounding effects of team-level adoption are most significant. An engineering team that has genuinely adopted Claude Code and Cursor ships at a fundamentally different pace. This is where CLOs should allocate the first and largest tranche of AI training budget.

2. Product Management (high urgency)

AI tools that let PMs prototype, validate, and spec faster compress the discovery-to-delivery cycle. PMs who can use Lovable or v0 to create clickable prototypes, Claude to draft and critique PRDs, and AI research tools to synthesise user data are dramatically more effective. The gap between AI-augmented PMs and traditional PMs is widening quickly.

3. GTM — Marketing, Sales, SDRs (high urgency)

The leverage in GTM comes from content velocity, personalisation at scale, and research acceleration. Marketing teams using AI for content production, SEO, and campaign ideation can 3–5× their output. Sales teams using AI for prospect research, call prep, and follow-up personalisation shorten deal cycles. These are among the fastest-moving lagging indicators post-training.

4. Operations and Finance (medium urgency)

The leverage here is in workflow automation — n8n and similar tools for connecting data sources, automating reporting, and eliminating repetitive manual processes. The upskilling need is real but tends to be narrower and more specific. Typically suited to internal build or a targeted buy for specific workflow automation skills.

What to look for in an AI training provider

The market is crowded and variable in quality. Evaluate providers on these criteria:

Curriculum freshness

Ask the provider when their core curriculum was last meaningfully updated and how often. If the answer is "annually" or "we update it as needed," walk away. The AI tool landscape changes significantly every quarter. You need a provider that is updating curriculum continuously and can point to specific recent changes they made in response to tool updates.

Practitioner instructors

The best AI training is delivered by people who are actively building with these tools on real projects. Ask providers: "Who are the instructors, and what are they building right now?" A facilitator with a certification is not the same as a practitioner who deployed a Claude Code workflow last Tuesday and can explain what worked and what did not.

Cohort-based model with live sessions

Asynchronous e-learning for AI tools is largely ineffective because the hardest parts of AI tool adoption — calibrating trust, handling unexpected outputs, debugging broken workflows — require real-time guidance. The best programs use a cohort model with live sessions where participants work through real tasks and can ask questions of a practitioner in the moment.

Outcomes measurement built in

Any serious provider should have a clear answer to "how will we know if this worked?" that goes beyond completion rates and satisfaction scores. If the provider does not propose a baseline measurement approach before training starts and a post-training metrics review, that is a significant red flag.

Reference customers in your industry and function

AI training outcomes are highly context-specific. An engineering cohort program that produced great results at a consumer app company may not transfer well to a regulated financial services environment. Ask for references specifically in your function and industry, not just name-brand logos on a slide.

Red flags in AI training programs to avoid

• "We cover all major AI tools": Breadth over depth is the defining failure mode of bad AI training. A program that mentions 20 tools produces engineers who know the name of 20 tools. A program that goes deep on 3 produces engineers who use 3 tools daily.
• No baseline measurement: If a provider is not proposing to baseline your metrics before training starts, they are not confident in their outcomes — or they do not plan to be accountable for them.
• Facilitators who cannot answer "how would you handle X" live: Test this in the sales process. Ask a scenario question about a specific tool limitation or edge case. A practitioner instructor should be able to answer it without hesitation. A script reader will struggle.
• Content that could have been written 18 months ago: Review the curriculum. If it mentions tools and capabilities that are already outdated, the provider is not keeping up. In AI training, stale content is not just inefficient — it actively builds the wrong mental models.
• No cohort or peer-learning component: Individual AI training produces individual habits. Team-level AI capability requires cohort-based learning where engineers (or marketers, or PMs) build shared mental models together.

How to build the internal business case for AI training budget

The strongest AI training business cases connect investment to specific, quantifiable workflow costs. Here is the structure:

• Identify the 2–3 highest-cost workflow bottlenecks AI can address: Use data from engineering managers, team leads, or your own L&D audit. Be specific: "Our engineering team spends an estimated 15% of total capacity on code review cycles that AI pre-review could compress by 30%."
• Calculate the conservative recovery value: Fully-loaded cost per hour × hours recoverable per person per week × headcount × weeks. Even conservative estimates typically produce eye-opening numbers at team scale.
• Present the risk cost of inaction: The competitive risk of having AI-naive teams while competitors upskill is real and increasing. Frame this concretely: what features are you not shipping? What campaigns are you not running? What operational efficiency are you leaving on the table?
• Propose a time-boxed pilot: Rather than asking for full-year budget upfront, propose a single-cohort pilot with defined success metrics and a 90-day review. This reduces the perceived risk of the investment while still getting the program started.

The 6-month roadmap for org-wide AI adoption

Org-wide AI adoption does not happen all at once. Here is a sequenced 6-month roadmap that balances speed with sustainability:

• Month 1 — Audit and prioritise: Conduct a cross-functional AI capability audit. Survey each team on current tool usage, perceived skill gaps, and highest-value workflow opportunities. Map findings to business priorities. Identify internal champions.
• Month 2 — Pilot cohort (highest-priority function): Launch the first external training cohort for your highest-leverage function (typically engineering). Establish baseline metrics before the cohort starts. Activate internal champions as cohort peers.
• Month 3 — Pilot review and expand: Review pilot metrics. If leading indicators are strong (daily usage, playbook entries), expand to the second-priority function. If they are not, diagnose and fix before scaling.
• Month 4 — Second cohort launch and internal build begin: Second cohort starts for the next function. Internal team begins building foundational AI literacy content (use policy, principles) for org-wide deployment.
• Month 5 — Champions community of practice launch: Formalise the internal AI champions network. Monthly community of practice sessions where champions share what is working, surface tool updates, and connect across functions.
• Month 6 — Full metrics review and Year 2 planning: Compile lagging indicators from pilot and second cohort. Build the Year 2 AI capability roadmap based on data, not assumptions. Present to C-suite with ROI evidence.

The 6-month roadmap is not a waterfall. Things will change — tools will update, team priorities will shift, a new model will drop and change the landscape. Build flexibility into the plan from the start. The goal is a learning organisation that can adapt to AI change, not a training program that assumes it knows exactly what AI will look like in 6 months.