How has hiring changed in this new world of AI confusion, panic, and opportunity.
The hiring mistake nobody talks about
Every business adding AI in 2026 is making the same hiring mistake.
They look at the problem (we should be using AI properly), see the size of it (this is going to touch every team), and reach for the familiar answer (we need to hire someone full-time to own this).
So they post a job. "Head of AI." "AI Transformation Lead." "VP of AI." They get hundreds of applications. They interview for six weeks. They hire someone with a strong CV and unclear remit. Six months later, the person has built a Notion page of "AI initiatives," run two pilots, and quit because nobody knew what they were supposed to do.
The mistake is not the hire. The mistake is the assumption that one person, full-time, is the right shape for the work.
It isn't. Not for most businesses. The work has three phases, and each phase needs a different operator.
Why AI work splits into three phases
Bringing AI into a business is not one job. It is three jobs that happen in sequence and then loop.
Audit. Before you build anything, somebody needs to look at how work actually happens. Where are the repetitive tasks? Where is the data trapped in PDFs and email threads? Which workflows are begging for automation and which ones are fine? This is a diagnostic job. It needs someone who has seen a lot of businesses, can pattern-match fast, and is honest enough to tell you that most of what you think needs AI doesn't.
Build. Once you know what to fix, someone needs to actually build it. Wire up the tools. Write the prompts. Set up the integrations. Train the model on your data. This is an engineering job. It needs hands on keyboards, not strategy decks.
Embed. A built system that nobody uses is worse than no system. Embedding means training the team, writing the SOPs, sitting next to the person who is going to use it every day until they trust it. This is a change-management job dressed up as a technical one.
Three phases. Three different skill sets. Three different time commitments.
The audit takes two to four weeks. The build takes one to three months. The embed is ongoing, low-intensity attention for six to twelve months.
No single person is great at all three. And no business needs all three of them full-time at the same time. Which is why the right shape for this work is fractional, sequenced, and matched to the phase.
A field guide to the operators
This is where the language gets messy. The new wave of AI and growth roles has produced a vocabulary that most buyers do not recognise yet. Here is the honest version of what each role actually means in 2026, including the ones I think are useful and the ones I think are noise.
Roles that matter for adding AI to a business
AI consultant / AI strategist. Does the audit. Comes in, looks at your workflows, tells you what to automate and what not to. Should leave behind a clear roadmap, not a slide deck. The good ones have shipped real systems before and will tell you when AI is the wrong answer. The bad ones are repackaged management consultants who learned to spell GPT.
Forward deployed AI engineer (FDE). Does the build, embedded with your team. The term comes from Palantir, where forward deployed engineers were sent to live with customers and build inside their workflows rather than handing over a product and walking away. It has since been adopted by AI-native companies including OpenAI and Anthropic. For a business adding AI, an FDE is the person who sits inside your operation for eight to twelve weeks, learns how you actually work, and ships the integration that fits your reality. Not a contractor handing over a repo. Someone who is in your Slack.
Applied AI engineer. Builds on top of foundation models. Prompt design, retrieval systems (RAG), agent workflows, evals. Distinct from a machine learning engineer, who trains models. Most businesses do not need ML engineers. They need applied AI engineers.
AI implementation specialist. Does the embed. Trains your team, writes the SOPs, handles the change management. This is the role with the least settled terminology - you will see "AI enablement lead," "AI adoption specialist," "AI change manager," all describing roughly the same job. Pick one and stick to it. We call it implementation.
Fractional CTO. For businesses without internal tech leadership, this is the person who owns the AI roadmap at a strategic level, picks the vendors, manages the FDEs, and stops you from getting sold something stupid. Usually two to four days a month.
Roles that matter for scaling a tech company
Fractional CMO / CRO / VP Sales / Head of Growth.
Senior commercial leadership, part-time. Established model that has matured significantly over the last five years. You hire one when you need the judgment of a senior operator but cannot justify a full-time cost. Common at pre-Series A and bridge stages.
Growth engineer.
An engineer embedded on a growth team. Ships experiments, builds internal tools, instruments funnels, automates the unsexy plumbing that makes growth work. The term has been in use since around 2021 and is well-understood inside Series A and B SaaS companies. Distinct from a marketing engineer (closer to MOps) and from a growth marketer (no code).
Fractional GTM (go-to-market).
Umbrella term for fractional commercial talent - usually a senior operator who can run sales, marketing, and partnerships at once for an early-stage company. The right hire when you have product-market fit signal but no commercial engine.
Roles I would treat with caution
AI coach. Could mean anything from executive coaching about AI strategy, to a prompt-engineering tutor, to a change-management consultant. If you hire one, get the scope in writing.
Prompt engineer. Real role two years ago. Increasingly absorbed into "applied AI engineer." Hire for the broader skill, not the narrow one.
AI transformation lead. Often a strategy job in disguise. Ask what they have actually shipped.
Why fractional, why now
Two forces are pushing this work toward fractional structures rather than full-time hires.
The first is velocity. Foundation models, tooling, and best practices are changing faster than any single full-time hire can keep up with. A fractional operator working across five clients sees five times the surface area. They bring back patterns. They know what is working at a company down the road that is solving the same problem you are.
The second is shape. The work genuinely is not full-time for most businesses. A small services firm does not need a full-time AI engineer. It needs a few weeks of audit, a couple of months of build, and then light-touch embed support for the next year. Force that into a full-time hire and you are paying full salary for a fraction of the work.
Fractional is not a discount on full-time. It is a different shape that happens to fit the work better.
How to actually do this
If you are a business owner or operator looking at your AI roadmap and trying to figure out what to do, the honest sequence is:
- Start with an audit, not a hire. Two to four weeks. Someone external, ideally fractional. Output is a list of three to five things worth automating, in priority order, with rough effort estimates.
- Build the highest-leverage one first. Bring in a forward deployed AI engineer or applied AI engineer for the build. Eight to twelve weeks, embedded.
- Embed before you build the next one. This is the part everyone skips. The build is half the work. Getting the team to actually use it is the other half.
- Then loop. Audit the next opportunity. Build. Embed. Each cycle is faster than the last because the foundation is there.
The businesses getting this right in 2026 are not the ones with the biggest AI teams. They are the ones who figured out the sequence and matched the right operator to each phase.
Where Rafiki fits
Rafiki Works is a fractional talent marketplace for exactly this shape of work. We match businesses with vetted operators across three engagement tiers - Audit, Build, Embed - covering AI consultants, forward deployed AI engineers, applied AI engineers, AI implementation specialists, and fractional commercial leadership for tech companies.
If you are starting an AI rollout, we will run a short audit and give you a roadmap. If you have the roadmap and need to build, we will match you with an FDE who has shipped this before. If you have built and need to embed, we will put an implementation specialist in your team at the right intensity.
FAQ
What is a forward deployed AI engineer?
A forward deployed AI engineer (FDE) is an engineer who embeds with your business and builds AI systems inside your actual workflows, rather than delivering a finished product from outside. The term originated at Palantir and has been adopted by AI-native companies including OpenAI and Anthropic. FDEs typically engage for eight to twelve weeks and work alongside your team in your tools.
What is the difference between an applied AI engineer and a machine learning engineer?
An applied AI engineer builds on top of existing foundation models using techniques like prompting, retrieval (RAG), agent workflows, and evals. A machine learning engineer trains models from scratch or fine-tunes them on specific data. Most businesses adding AI need applied AI engineers, not ML engineers.
What is a growth engineer?
A growth engineer is an engineer embedded on a growth team who ships experiments, builds internal tools, and instruments funnels. The role has been established inside Series A and B SaaS companies since around 2021. Growth engineers are distinct from growth marketers (no code) and from marketing engineers (closer to marketing operations).
What does fractional mean?
Fractional means part-time, senior, and ongoing. A fractional operator works with multiple clients across a week or month, bringing senior judgment without the cost of a full-time hire. Common roles include fractional CMO, fractional CTO, fractional CRO, and fractional head of growth.
What is the difference between Audit, Build, and Embed?
Audit is the diagnostic phase: looking at workflows to identify where AI fits and where it does not. Build is the engineering phase: shipping the actual integrations and systems. Embed is the adoption phase: training teams, writing SOPs, and ensuring the built systems are actually used. Each phase needs different talent and different time commitments.
Do I need a full-time AI hire?
Most small and mid-sized businesses do not. The work splits into three phases (audit, build, embed) with different intensities and different skill sets. A sequenced fractional approach typically delivers better outcomes at lower cost than a single full-time hire, particularly in the first year of an AI rollout.
For all FAQs on fractional talent, compliance and payments, visit our full Rafiki Works FAQs page here.
