So… you’re building an agentic AI product or feature? Looking to hire agentic AI developers, architects or product managers? Feeling confused by the market, the talent pool, and what an ideal candidate’s background & experiences should look like?
Don’t worry, you’re not the only one.
In the last few months, the Artemis team has helped hire ~6 roles requiring agentic AI experience, and spoken to >50 candidates with this expertise. One thing is clear: Hiring for Agentic AI talent (and understanding what that talent can actually do for your org) is top of mind right now.
From our hiring perspective to yours, here’s what we’ve learned.
Let’s dig in.
First things first, what is the state of Agentic AI talent right now? Here’s what we found:
- Production vs. beta: Many companies claiming to have deployed agentic AI only have it deployed in beta with limited customer exposure
- Internal tooling vs. customer-facing: The talent pool of engineers who’ve built and deployed agentic tools for customer-facing products - with all of the complexity it comes with like security, observability, and traceability - is much smaller than those who’ve deployed internal tools
- The shhh factor: Teams with real production deployments often aren’t talking about it. Why? Our guess is to hide their talent from recruiters and to keep competitors guessing.
- A limited (but growing!) talent pool: Three months ago, production-ready, customer-facing agentic AI tools were pretty rare in Canada: Most teams we spoke to were still in beta or pre-production. In our latest search, we’re seeing more teams actually move into production. It’s still relatively early, but the landscape is changing fast.
Let’s get clear on what true Agentic AI is and isn’t:
- Depends on who you ask: Some organizations are referring to RAG (Retrieval Augmented Generation) or simple workflow automations as "agentic AI" while others refer to planning and executing multi-step workflows, orchestrating tools, reasoning through problems, and iterating independently to get to the right outcome. For our purposes, we were looking for the latter.
- At the very least, more than a Chatbot: In our conversations, we found that many deployed “agentic” solutions were really just traditional chatbots. They serve a purpose, but not the complex logic, reasoning and orchestration we’re talking about when it comes to this talent.
- Which means… high risk, high reward: The companies with truly autonomous agentic features need to protect themselves from agents going off track. They need far more sophisticated security + guardrails, evaluations and observability frameworks.
Ok, but which companies are really going for it, and how?
- Big orgs are spending: Large enterprises like Oracle and Shopify have the cash - and are splurging - when it comes to agentic AI product development
- Hire-and-fire: In our conversations, we found one larger company that hired a team with agentic skills, launched their solution, and promptly laid off most of the agentic team. Not ideal. Don’t do that.
- The birth of Agentic-AI-Native startups: We found a number of startups with agentic solutions as their core solution, as opposed to layering it on top of existing non-AI products.
- That technical debt buried in your basement? About that…: Many established software orgs are adding agentic layers to their existing enterprise platforms, working in parallel on core platform improvements and crossing their dear fingers that one doesn’t break the other.
In our conversations, we found the organizations successful with this are the ones who seriously committed to fixing their technical debt, ie the code with the cobwebs in the closet, y’know?
Now for the fun part – the talent pool. Here’s the Artie-tea:
- Years of experience can’t be a measure of skillset: Agentic AI is new-new. The most experienced talent has one year of true experience in this particular skillset, tops. Instead of years of experience, you should be looking for: (1) Agentic and platform complexity, (2) End to end - or at least major - ownership of the agent and infrastructure, (3) A keen understanding of the security required for the product’s use-case and how to mitigate the right risks.
- Two flavours of AI Talent: Data science and machine learning AI talent tends to come from academia and research institutes. On the other hand, some experienced software developers are focusing on agentic AI and building the systems in which agents operate. Oftentimes, the agentic AI teams require both working in synergy.
- The market is hot-hot-hot… and the talent pool knows it: Normally, our technical senior individual-contributor searches require high-volume outreach because the candidate pool is less responsive. In contrast, our agentic AI candidates are highly engaged, shopping for the best compensation, and sitting on multiple, competing offers.
- Speaking of, prepare to cough up some cash: In general, we found that candidates coming directly from academia or research without much developer experience were expecting $160-190K base. At the other end of the spectrum, senior software devs with agentic AI experience are asking for >$200K base (at the staff/principal level add 40-50%).
Speaking of motivations, what are agentic AI developers looking for to make a move?
- Compensation: Enough said. But as always be wary of candidates who are solely financially motivated. It is fair to ask for what you are worth, but test a candidate’s likelihood to hop to a new role when more $$ is offered.
- Full lifecycle ownership: Candidates want the opportunity to make fundamental decisions about the agentic AI frameworks, to prototype and also work on the production side of the SDLC, not just hold “a slice” of the agentic system.
- Mentorship: Given the stakes, speed and evolving nature of agentic AI, candidates are really drawn to opportunities with experienced agentic architects. This talent will cost a lot… but the payback in talent attraction and retention for the rest of your agentic AI team will be tangible.
- Project novelty: Similar to above, simple information retrieval chatbots aren’t compelling for candidates. Top talent wants to push the boundaries of what’s possible with autonomous agents.
That’s what we’ve got for now, friends of Artemis. But with the volume of these Agentic AI searches we’re being asked to do, you bet the learnings will keep coming.
Until next time,
- Negin, Tara & The Artemis Team
.jpg)
.jpg)
