By the Live Assets Team, with insights from Olga Fragis, Founder & CEO
A year ago, the most valued AI skill was simple: use it as much as possible. Companies put employees on leaderboards to see who could burn the most tokens. Heavy usage was a badge of productivity. Then the bills arrived. And almost overnight, the definition of a strong AI hire changed completely.
Something quietly shifted in the technology market over the last few months, and it is already changing what our clients ask us to find.
For most of 2025, the message from leadership was clear: adopt AI everywhere, experiment freely, do not worry about the cost. The push was so intense it earned a name. Tokenmaxxing. Companies like Meta and Amazon reportedly ran internal leaderboards where employees competed to consume the most AI tokens. Heavy usage became shorthand for being productive.
Then the invoices came due.
Uber reportedly spent its entire 2026 AI budget in the first four months of the year. OpenAI’s CEO called AI costs a “huge issue” for customers almost overnight. Companies that had encouraged unlimited experimentation started imposing per-employee spending caps. The Linux Foundation launched a whole new standards body, the Tokenomics Foundation, just to bring cost discipline to AI spending the way FinOps did for cloud.
The era of “use AI at any cost” slammed straight into the era of “who approved this invoice?”
For anyone hiring IT talent, this is not background noise. It is a fundamental shift in what a strong AI hire looks like.
Why “Strong AI Skills” Means Something New in 2026
Here is the change in one sentence. It is no longer about how heavily a candidate uses AI. It is about how efficiently they apply it to deliver measurable outcomes.
That sounds subtle. In practice, it completely reframes the hiring conversation.
A year ago, a candidate who said “I use AI for everything” sounded impressive. Today, a sharp hiring manager hears that and wonders what it is costing. The engineer who genuinely stands out now is the one who knows when to reach for a large model, when a small one will do, and when not to use AI at all. The one who can look at a workflow and say, “we are making five API calls here when one would do.”
This is the same pattern we have seen before. Every new computing wave starts with unlimited enthusiasm and ends with cost discipline. Cloud went through it. Mobile went through it. AI is just moving through the cycle faster than anything before it.
We wrote recently about hiring AI talent in Toronto, and the roles dominating that market. This is the natural next chapter. Companies do not just need AI talent anymore. They need cost-aware AI talent.
“The clients who were saying ‘find me anyone who knows AI’ last year are now asking very different questions. They want to know if a candidate understands what AI actually costs to run, and whether they can build something efficient. That is a real skill, and right now it is in short supply.”
Olga Fragis, Founder & CEO, Live Assets IT Staffing Solutions
The 5 AI Skill Areas Worth Hiring For Now
Based on what we are hearing from clients across Toronto and North America, here are the five skill areas that have moved from “nice to have” to “genuinely valuable” in the space of a few months.
1. AI Cost Literacy
The single most in-demand emerging skill. Cost literacy means a candidate genuinely understands the economics of what they are building. They can read token and API cost structures across large language models, agents, and tools. They can estimate the cost per task, whether that is coding, summarisation, or data extraction. They can weigh model cost against performance and choose the cheaper option when high accuracy is not required.
This is the AI equivalent of an engineer who understands cloud billing. Rare, valuable, and increasingly non-negotiable for any role touching production AI systems.
2. Prompt Efficiency and Structured Thinking
Anyone can write a prompt. Far fewer people can write a concise, high-signal prompt that gets the right answer without burning tokens on the way. The candidates worth hiring know how to avoid repeated or recursive AI calls, structure prompts for deterministic outputs, and reuse templates instead of improvising every time. Small habits, but at production scale they translate directly into the monthly bill.
3. AI Workflow Optimisation
This is where real money is saved or wasted. The skilled candidate designs multi-step workflows that minimise AI calls. They know when not to use AI at all, combining rule-based logic with AI selectively rather than reaching for a model on every step. They can spot an agent stuck in an unnecessary loop and fix it. Per-developer token consumption reportedly rose nearly 19 times in nine months, much of it from autonomous agents running unchecked. The engineers who can rein that in are worth their weight.
4. Model Selection and Orchestration
Not every task needs the most powerful model. A strong AI hire in 2026 knows how to select small versus large models appropriately, build hybrid architectures that combine rules, retrieval, and language models, and route tasks based on complexity. They understand the trade-offs between latency, cost, and accuracy, and they make those calls deliberately rather than defaulting to the biggest, most expensive model every time.
5. AI Observability and Usage Monitoring
You cannot manage what you cannot see. This skill set covers tracking token usage by team, feature, or function, monitoring cost spikes in AI workflows, and setting budgets, alerts, and throttles. It also includes analysing the actual return on investment of AI features against the quality of their output. As AI spend becomes a board-level concern, the people who can make it visible and accountable become essential.
This Is Not Just an Engineering Problem
One important nuance our clients are learning quickly: cost-efficient AI is not owned by a single department. It is a shared responsibility, and the smartest organisations are hiring with that in mind across several functions.
Architecture sets the design rules that determine how efficient a system can be. Engineering controls execution efficiency at the code level. Platform and SRE teams enforce and monitor the cost guardrails. Product and delivery validate whether the AI feature is actually delivering business value. And finance, increasingly through dedicated FinOps roles, governs the overall spend.
For hiring leaders, this means the AI cost conversation is not a niche you can solve with one specialist. It is a capability you build across the team, from the architect down to the FinOps analyst. The companies getting this right are hiring for cost-awareness as a thread that runs through multiple roles, not a box ticked by a single hire.
What This Means for Your Hiring Strategy
If you are building or scaling a team that touches AI in any meaningful way, here is the practical takeaway. The interview questions need to change.
Instead of asking “how do you use AI,” ask “how do you decide when not to use AI.” Instead of “what models have you worked with,” ask “how do you choose between a small model and a large one for a given task.” Instead of “have you built AI agents,” ask “how do you keep an agent from running up unnecessary costs.” The answers will separate the candidates who chased the tokenmaxxing trend from the ones who actually understand the economics of what they are building.
This connects to something we believe deeply. The best hires have always gone beyond the keywords on a resume. A candidate can list every AI tool under the sun and still have no idea what any of it costs to run at scale. The skill that matters now is judgment, and judgment does not show up in a checklist of technologies.
Hiring AI talent and not sure what skills actually matter anymore?
Whether you are building an IT team, rethinking your hiring process, or just want a more human approach to recruitment, we would love to have a real conversation. No pitch decks. No pressure. Just two people figuring out what is next.
Get in touch today, here.
Discussion