As part of The State of the AI Frontier 2026, AI Frontier Network invited leaders building and deploying AI in the real world to share where the frontier is actually moving. In this contribution, Agnese Jaunosane, Co-founder & COO at Ajelix, gives a candid read on what changes in 2026 — and what to watch.
On the shift that defines 2026
Between now and the end of 2026, the most significant change is that agentic AI moves from demo to daily infrastructure, and the economic assumptions built around AI are starting to break.
Adoption of AI agents is accelerating faster than organizations anticipated. But alongside that, we're watching unit economics shift. Token costs are rising, the SaaS subscription model is under real pressure to move to consumption-based AI pricing, and new business models are emerging to compensate. Platforms that assumed flat monthly fees are rebuilding their pricing architecture.
At the same time, a regional split is hardening. The EU is actively seeking open-source alternatives that can be hosted locally and audited transparently, while US-based labs like Anthropic are publicly arguing that open-source models present safety risks. That tension is going to intensify. The organizations caught in the middle are asking a question that wasn't on their roadmap a year ago: can we continue to depend on a handful of US-based providers for critical infrastructure?
For many enterprise buyers, the answer is increasingly no. Self-hosting is moving from an edge case to a governance requirement. We expect more capable open-source models to launch in this period, and the pressure to run models on owned infrastructure, especially following platform shutdowns like Fable, will only grow. Companies are trying to maintain AI capability while managing concentration risk. That balance is the defining operating challenge of the second half of 2026.
On the real unlock
There's a persistent mental model in the industry that the only serious options are a small number of large AI proprietary providers. That model is slowly shifting. A wave of capable open-source models and agent frameworks now exists that can be fine-tuned, self-hosted, and deployed in production, at a fraction of the cost of proprietary APIs, and with full control over data.
The unlock is the combination of model capability with the infrastructure to run it in a way that meets enterprise requirements around security, compliance, and cost predictability. Most organizations haven't yet adapted their thinking or their procurement process to treat open-source as a serious operational option. The companies that make that shift in 2026 will have a meaningful structural advantage over those that remain dependent on external providers. The technology is ready. The adoption mindset hasn't caught up.
On the trap to avoid
The most dangerous bet leaders are making in 2026 is treating AI as a solution before they've understood the problem.
The specific failure mode: reaching for an AI agent or a complex workflow when a simple, deterministic automation would do the job ten times cheaper and more reliably. We see this constantly organizations that come to us wanting multi-agent orchestration for something that would be better served by a basic conditional logic flow. AI is the right tool for tasks that are ambiguous, variable, or require judgment. It is the wrong tool for tasks that are predictable and structured.
The deeper version of this trap is believing that AI can substitute for process thinking entirely. A poorly designed workflow fed into a great AI system still produces poor outputs. The organizations that will extract real value from AI in 2026 are the ones with humans who understand what good looks like, people who can assess what the AI delivered, identify the 10% that needs correction, and close the gap. That expertise is not something AI can replace. What we're seeing instead is that many leaders are investing heavily in tools while underinvesting in the judgment needed to use them well. The magic pill framing is still everywhere, and it's costing organizations real money.
On the organization that adapts
AI is forcing a hard reset on how we think about junior and senior roles.
We gave our junior team members AI tools expecting faster delivery and better output. What we found instead was that they were outsourcing their thinking, not just their execution. They lacked the baseline experience to evaluate what the AI produced. They couldn't tell when an output was 90% right versus fundamentally wrong. We've had to go back and rebuild the habit of critical first-principles thinking before they touch an AI prompt. For juniors, AI without judgment is a liability, not an accelerant.
For senior contributors, the opposite is true. Experienced people already carry a clear model of what the output should look like, they can specify what they need precisely enough for AI to deliver it, and they can validate the result in minutes. We no longer need junior technical staff for work that a well-directed agent can execute. That capacity has been reallocated entirely to marketing and go-to-market, which is the one area where AI genuinely cannot substitute for human judgment and creativity.
The time we previously spent explaining requirements to juniors, we now spend explaining requirements to AI. The difference is the AI ships the first version in minutes and we close the last 10%. That's a real productivity shift, but only because senior judgment is still driving it.
On the benchmark that matters
The metric that will define AI's trajectory in the second half of 2026 is ROI per dollar of AI spend.
Whether the investment is actually generating measurable return.
We are already seeing the early signals of a reckoning. Enterprise organizations are running significant AI budgets with unclear payoff timelines. Microsoft's decision to pull back on Claude Code subscriptions and Uber's AI cost overruns are early data points in what will become a broader pattern of AI budget scrutiny. For platforms like ours, ROI shows up in every enterprise conversation. "Adoption is interesting, but what's the business case?" is now the opening question, not an afterthought.
By December 2026, we expect ROI to be the filter through which most enterprise AI investment gets evaluated, and we expect a meaningful number of current AI initiatives to fail that test. The companies that ship AI with clear, measurable impact on specific workflows will consolidate share. Everything else will face hard questions in annual planning cycles.
