On October 1, 2026, The Conference Board published its CHRO Confidence Index for the third quarter. On the surface, CHROs remain in positive territory, though confidence continues to inch down. Fifty-five percent of chief human resources officers consider their organization at least somewhat prepared to manage AI-driven workforce change over the next two years, and 53% expect hiring to increase over the next six months.
Look one level down and the confidence thins. Only 4% say they are very prepared, and 27% say their organization is unprepared. More than eight in ten (82%) say their organization has changed its workforce-planning process because of AI, but only 7% report a significant change. Just 19% say anticipated AI effects have been incorporated into workforce and financial planning across the enterprise.
The CHROs also named what is holding them back: difficulty predicting AI's impact on the workforce (41%), limited AI skills (39%), and an unclear AI strategy (31%). Two of those three are planning problems rather than technology problems, and limited AI skills is the one a workforce plan can address most directly in the year ahead.
That is the gap this piece is about. Most companies know AI will change the skills they need in 2027. Far fewer have decided how they will get those skills: by building them in the people they have, buying them through new hires, or borrowing them through contractors and partners.
A disclosure before we go further. Golabs provides nearshore AI staffing and dedicated teams, which is the "borrow" option, and we run an AI training program, which is part of "build." So we have a commercial interest in two of the three answers. We have tried to rely on nonprofit research and analyst sources, and we flag the commercial interest of others where it applies.
Should You Build, Buy or Borrow AI Skills in 2027?
Most organizations will need all three, assigned seat by seat. Build where the skill sits close to your business, and people already use AI daily. Buy where you need permanent, senior AI expertise you will own. Borrow where the need is urgent, specialized, or uncertain. Only 19% have built AI into enterprise plans so far.
Build means upskilling the people you already employ. In our experience, it costs the least per person and takes the longest to produce deep expertise. It also keeps business knowledge in the building, because the people learning AI are the ones who already understand your customers, your processes, and data.
Buy means hiring permanent employees with the AI skills you need. It gives you ownership and continuity, at a higher price, and senior AI hires can take months to land. It is the right tool for the few seats that will define your AI capability for years, and an expensive one for everything else.
Borrow means bringing in skills through contractors, staff augmentation, or a partner team. It is the fastest to start and the easiest to scale down, and it requires deliberate knowledge transfer if you want the capability to stay. Without that transfer, the work ships but the know-how leaves with the team.
A common mistake we see in our own recruiting work is treating this as one company-wide decision. It is a decision for each seat. A single team might train its analysts, hire one permanent architect, and borrow two senior engineers for a first build, all in the same quarter.
What Does It Take to Build AI Skills In-House?
More than access to a tool. The usage is already there; what is missing is structured training and the time to take it. The Conference Board's July 2026 study on AI skilling, based on interviews with 35 enterprise leaders and a global survey of nearly 1,300 workers, found that 55.1% of workers use generative AI or AI agents daily or weekly.
Yet only 33.3% had used organization-provided AI training in the past six months, and 28.3% said their organization does not provide AI training at all. Only 48% agreed their organization gives them sufficient time during work hours for AI skills development, and only 47.6% said they had sufficient tools, access, and resources.
Pay systems point the same way. In The Conference Board's US salary increase budget survey, published September 3, 2026, 38% of organizations said AI and machine-learning skills are driving base-pay increases, but just 3.4% specifically reward hot-skill acquisition, and 63% still base salary ranges entirely on an employee's job rather than skills. AI skills move pay, but few companies reward acquiring them.
Build is the right answer when the skill depends on knowing your business: AI-assisted work in finance, operations, customer service, and product, where context matters more than deep technical expertise. Our earlier guide on preparing your workforce for AI adoption covers the rollout side, from choosing first use cases to setting expectations with managers.
Build is the wrong answer when you need someone to design, ship, and operate production AI systems next quarter. In our recruiting work, training a generalist engineer into that role takes longer than most roadmaps allow. Build is how you grow that capability over time, not how you get it by a deadline.
What Does It Cost to Buy AI Talent?
More than the average raise suggests, and the competition is not easing. The Conference Board projects the median salary increase budget to stay at 3.5% in 2027, with technology, telecommunications, and information services at 3.7%. That budget covers raises for people already on payroll. New AI hires are priced differently.
We broke down the 2027 numbers in what you should pay AI engineers. Robert Half, a staffing firm, estimates 2027 starting salaries for an AI engineer to range from $154,500 to $196,750. Those are starting salaries, before recruiting costs, which is why each buy decision deserves real scrutiny.
Demand keeps rising. CompTIA's analysis of September data found 349,821 active job postings requiring AI-related skills, up 32,327 from August, and 625,633 active tech postings overall, the highest total in more than three years. In the same month, employment in tech occupations across all industries fell by roughly 6,000 positions.
Tech unemployment edged up to 3.3%. CompTIA sells certifications, and one month is not a trend. Still, in September employers posted more AI roles while tech employment dipped. We read that as a gap in specific skills rather than in headcount, which is exactly the kind of gap a buy decision is meant to close.
There is also a warning for companies tempted to cut first and rehire later. Gartner predicted on September 9, 2026, that by 2029, 30% of employees laid off due to AI-driven replacement will need to be rehired, often at a significantly higher cost. That is a prediction, not a measurement, but it deserves attention.
It fits the expectations gap in McKinsey's 2026 State of AI survey of 1,719 participants: 14% of respondents from organizations using AI said it contributed to an overall decline in workforce size over the past year, while 39% expect AI to decrease their organization's headcount over the next year. Expectations are running ahead of experience.
Buy is the right answer when the seat is permanent, senior, and central to your AI strategy: the architect who will own your AI platform, or the engineering lead who will be accountable for it in three years. These are the people whose judgment you want inside the company when the strategy changes.
Buy is the wrong answer when you are hiring permanently for a need you cannot yet size. The CHROs above told The Conference Board their top barrier is predicting AI's workforce impact. A permanent hire is an expensive bet on a forecast you do not trust yet.
When Does It Make Sense to Borrow AI Talent?
When speed matters, the need is specialized, or the size of the need is still uncertain. Robert Half's technology salary research notes that many organizations are turning to contract talent for roles in AI governance, automation, and cloud initiatives. That matches what we see. Borrowing fits three situations.
A project with a clear end. Building a first agent, migrating a data pipeline, or standing up evaluation tooling has a beginning and an end. Hiring permanently for a nine-month project leaves you with a role to redesign afterward. Borrowing lets the role end when the project does.
A skill you need before you can train it. If nobody on your team has shipped a production AI system, borrowing a senior team can deliver the first system and teach your people in the process. Done well, borrow becomes build. The key is pairing your people with the borrowed team from the first week, not at handover.
A need you cannot size yet. When 41% of CHROs say they struggle to predict AI's impact on the workforce, flexible capacity buys time to learn what you actually need before committing to permanent headcount. Once the picture is clearer, you can convert the right seats to build or buy.
Borrowing has real risks. Knowledge can walk out the door with the contractors. Remote hiring has become a target for fraud, which we covered when vetting remote engineers in the deepfake era. And partner quality varies widely. Each risk is manageable, but only if you plan for it before the contract starts.
A caution about our own corner of the market. We know of no independent 2026 dataset on AI engineering supply, pay, or time to hire in Latin America comparable to the US sources above. Anyone quoting precise nearshore savings is quoting their own book of business, and that includes us. Ask any partner, us included, for rates and timelines in writing.
How Do You Put Build, Buy and Borrow Into One 2027 Plan?
List the AI work, not the AI roles. Start with the work your 2027 plan requires: which systems you will build, which processes AI will change, who will run what you deploy. Roles follow from the work. A role list drafted first tends to copy job titles from other companies rather than describe what you need.
Sort every seat into build, buy, or borrow. For each one, ask three questions: does it need deep knowledge of our business, will we need it in three years, and how soon do we need it at full strength? Business knowledge points to build, permanence points to buy, urgency and uncertainty point to borrow.
Fund the training you are counting on. If your plan assumes people will learn AI skills, budget the hours. Only 48% of workers in The Conference Board's study said they have enough time during work hours to develop AI skills. A training plan without time is not a plan. For structured help, our AI fluency program is one option.
Reward the skills you want to grow. If 38% of organizations say AI skills drive base pay but only 3.4% reward people for acquiring them, the incentive points outward. Recognize internal skill gains in pay and promotion, or expect your best learners to leave. The goal is for a new skill to count in decision-making.
Write knowledge transfer into every borrow contract. Set documentation, pairing, and handover requirements up front, so borrowed capacity leaves capability behind. A dedicated AI team works best with an exit plan from day one, and our AI staffing models guide compares the structures and when each one fits.
Revisit the mix every quarter. The Conference Board's numbers move each quarter, and so will your needs. A seat you borrow in the first quarter may be one you buy or build by the fourth. Put the first review on the calendar now, so it happens before next year's budgets lock.
Common Questions About Planning AI Skills for 2027
What does build, buy, or borrow mean in workforce planning?
Build means developing your current employees' skills through training. Buy means hiring permanent employees who already have the skills. Borrow means bringing skills in through contractors, staff augmentation, or a partner team. Most organizations planning for AI in 2027 will use all three, choosing the approach seat by seat rather than company-wide.
How prepared are companies for AI workforce change?
Not very, by their own account. The Conference Board's Q3 2026 CHRO Confidence Index found 55% of CHROs say their organization is at least somewhat prepared for AI-driven workforce change over the next two years, but only 4% say very prepared. Just 19% have built anticipated AI effects into workforce and financial planning across the enterprise.
Is it cheaper to upskill employees or hire AI talent?
In our experience, upskilling costs less per person but takes longer to produce senior AI engineering skills. Hiring brings specialized skills faster once the role is filled, but it is expensive: Robert Half's 2027 guide puts AI engineer starting salaries at $154,500 to $196,750. Many companies combine both, borrowing senior talent while their own people train.
When should a company use contract or augmented AI talent?
Contract or augmented AI talent fits projects with a clear end, skills your team needs before it can learn them, and needs you cannot size yet. It is fastest to start and easiest to scale down. To retain the capability, include documentation, pairing, and handover requirements in the contract from the start.
If you are building your 2027 plan and want a second opinion on which AI seats to build, buy, or borrow, talk to the Golabs team. We will start with the work, and we will tell you which seats you should not hand to a partner.