Budget conversations in 2026 are unusually consistent. Security spending is protected, data infrastructure is protected, and almost everything else has to justify itself against an AI initiative.
That pressure has produced a clear hierarchy of what gets funded. The useful thing for anyone hiring is that the hierarchy is visible in posting data and salary movement rather than in opinion pieces.
Below are the ten in-demand technology roles showing the strongest sustained demand, what each is actually being hired to do, and the trap in each one. Rise92 sources against this list for distributed engineering teams.
The ten roles, with 2026 US salary bands
Ranges from Robert Half’s 2026 Salary Guide, based on analysis of thousands of postings and placements. All figures are US national and in USD.
| Role | Low | Midpoint | High |
| AI/ML engineer | $134,000 | $170,750 | $193,250 |
| Data engineer | $127,000 | $156,250 | $180,750 |
| Data scientist | $121,750 | $153,750 | $182,500 |
| DevOps engineer | $118,000 | $145,750 | $173,750 |
| Cybersecurity engineer | $118,500 | $144,000 | $190,750 |
| Software engineer | $109,250 | $142,000 | $175,500 |
| Network and cloud engineer | $110,000 | $132,000 | $155,000 |
| ERP business analyst | $101,000 | $123,250 | $142,250 |
| IT project manager | $103,500 | $122,750 | $147,000 |
| Systems administrator | $80,250 | $98,000 | $118,000 |
Note the spread rather than the midpoint. Cybersecurity has the widest band of any role here, running from $118,500 to $190,750, which tells you seniority is valued far more steeply there than the average suggests.
Where the budget is actually going
Business leader priorities for 2026, per Robert Half’s research:
| Priority | Share of leaders citing it |
| AI and machine learning | 45% |
| IT operations and infrastructure | 36% |
| IT governance and compliance | 25% |
| Cloud architecture and operations | 24% |
| Data engineering and analytics | 22% |
The pattern worth noticing: AI leads, but the next four are all about running things reliably and compliantly. Technical hiring priorities in 2026 are split between building new capability and making sure the existing estate does not fall over while you build it.
1. AI and ML engineer
Why prioritized: the highest-paid role on the list and the clearest budget line. Roughly 41% of US tech job advertisements now require or focus on AI, with demand growing roughly sevenfold in two years.
What they are actually hired to do: put models into production and keep them working. Retrieval, evaluation, deployment, and the unglamorous maintenance underneath.
Screen for: a deployed system with real users, and a specific account of what broke and how they fixed it.
The trap: hiring research-oriented profiles for production work. The interview processes differ completely, and running the wrong one filters out exactly the people you needed.
2. Data engineer
Why prioritized: every AI initiative eventually stalls on data quality, and the organisation discovers the pipeline work was never done. Data engineering and analytics is cited by 22% of leaders as a top priority.
What they are actually hired to do: build and maintain the pipelines that everything else depends on, including ingestion, transformation, quality checks, and lineage.
Screen for: experience with pipeline failure at scale, not just tooling familiarity.
The trap: treating this as a support function. Data engineering roles are frequently the constraint on whether the AI programme ships at all, and staffing them last is a common and expensive sequencing error.
3. Data scientist
Why prioritized: still in demand, though the role is being actively redefined toward machine learning and production work rather than analysis and reporting.
What they are actually hired to do: increasingly, build things rather than produce insight decks. The boundary with ML engineering has blurred considerably.
Screen for: whether they have shipped a model into production, or only worked in notebooks. That single distinction separates two very different hires.
The trap: writing a job description for the 2020 version of this role and interviewing for the 2026 version, or the reverse.
4. DevOps engineer
Why prioritized: IT operations and infrastructure is the second-highest cited priority at 36%, and cloud specialties are projected to expand in the low-to-mid teens.
What they are actually hired to do: deployment pipelines, infrastructure as code, and increasingly observability for systems with non-deterministic components.
Screen for: the specific stack. A DevOps engineer who knows Terraform, Kubernetes, and can build observability pipelines is a completely different hire from one who writes Jenkins scripts.
The trap: the title covers such a wide capability range that job-title matching is close to meaningless here.
5. Cybersecurity engineer
Why prioritized: risk is rising faster than capacity. Projected growth sits around 29% to 35%, with close to half a million open positions in the US alone, and IT governance and compliance is cited by 25% of leaders.
What they are actually hired to do: threat detection, secure architecture, incident response, and increasingly the security review of generated code and agent systems.
Screen for: cloud security, zero trust architecture, DevSecOps, and framework experience such as ISO 27001 or SOC 2.
The trap: underpaying against the top of the band. This role has the widest spread on the list for a reason, and mid-band offers do not attract senior security people.
6. Software engineer
Why prioritized: the volume role. AI tooling changed what the job involves without reducing how many are needed.
What they are actually hired to do: the work has shifted toward review, verification, integration, and judgment rather than raw production.
Screen for: the ability to find what is wrong with code rather than to produce it quickly.
The trap: running a 2019 interview loop for a 2026 job. Algorithm puzzle performance predicts less about this role than it ever did.
7. Network and cloud engineer
Why prioritized: cloud architecture and operations is cited by 24% of leaders, and AI workloads for training and inference are deployed overwhelmingly in cloud environments.
What they are actually hired to do: hybrid environments, migration from legacy on-premise systems, and cost control on infrastructure that grew faster than anyone budgeted.
Screen for: cost optimisation experience specifically. Inference and storage costs at scale are usually larger than forecast.
The trap: conflating this with DevOps. Related, overlapping, not the same hire.
8. ERP business analyst
Why prioritized: the quiet entry on the list, driven by modernisation programmes that cannot be delayed and regulated industries still moving off legacy systems.
What they are actually hired to do: translate between business process and system capability during large transformations.
Screen for: whether they have survived a full implementation cycle, including the part where the original plan failed.
The trap: dismissing this as non-technical. It is frequently the role that determines whether a multi-year programme lands.
9. IT project manager
Why prioritized: organisations are running security hardening, AI implementation, and system modernisation simultaneously with limited capacity. Someone has to sequence that.
What they are actually hired to do: manage colliding priorities across teams that do not share incentives.
Screen for: experience with technical dependencies, not just delivery frameworks and certifications.
The trap: hiring for methodology fluency over technical judgment. Certification says almost nothing about whether someone can spot an unrealistic engineering estimate.
10. Systems administrator
Why prioritized: the lowest-paid role here and the most consistently underestimated. IT operations remains the second-largest budget priority, and someone runs the estate.
What they are actually hired to do: keep existing infrastructure working while everything around it changes.
Screen for: breadth across a heterogeneous environment rather than depth in one product.
The trap: assuming this role is being automated away. Entry-level and routine work is shifting, but the person who knows why the legacy system behaves that way is not replaceable by tooling.
The title problem
Every role above has a caption problem, and it is the main reason searches fail.
| Same title | Two completely different hires |
| DevOps engineer | Terraform, Kubernetes, observability pipelines / Jenkins scripts only |
| Data scientist | Production ML pipelines/notebook analysis only |
| AI engineer | Deployed and maintained a live system / built demos |
| Cybersecurity engineer | Secure architecture and incident response/compliance checklists |
The practical rule: hire for capabilities rather than titles. The title tells you the category. The skills tell you whether the person can do the actual work, and the gap between two people holding the same title is frequently larger than the gap between two different titles.
What this means for your job descriptions: specify the capability, not the label. A posting that says “DevOps engineer, five-year experience” attracts a pool where half cannot do what you need, and you will not find out until round three.
Why these roles are hard to fill
Three compounding constraints, and only one is about headcount budget.
Scarcity is concentrated, not general. Cloud-native engineering, generative AI work, zero-trust security, and advanced site reliability engineering are repeatedly flagged as scarce, high-complexity areas. There is no shortage of people with the title. There is a shortage of people who have operated at scale.
Operations roles are the hardest. Enterprise HR leaders have publicly flagged AI and ML operations roles as difficult to fill because qualified supply is limited while demand is strong. Building capability is better supplied than running it.
The strongest people are not searching. Senior professionals in all ten categories are employed, well compensated, and not reading job boards. Reaching them takes a network route, which is the basis of off-market sourcing rather than broader advertising.
The junior pipeline complicates this further. AI tooling is absorbing routine tasks previously handled by junior developers, QA testers, and IT support staff, which means fewer people are getting the supervised practice that produces senior judgment. Companies that stop developing juniors will be competing harder for seniors in five years.
What this costs, and where the supply is
The salary table above is US-centric. The same capabilities exist in other markets with substantially different cost structures, which changes the arithmetic on team composition rather than on quality.
For instance, Pakistan market data counts 300,000+ export-ready engineers across AI, cybersecurity, blockchain, and cloud, within a pool of 2.3M+ digital professionals. The honest qualification is that senior production depth in AI specifically is thinner than headline figures suggest, so sourcing method matters more in that category than in cloud or data engineering.
Two things determine whether the economics actually work. The sourcing has to reach genuine senior depth rather than the actively searching tier, and the employment structure has to be compliant in the relevant jurisdiction. Both are solvable, and the full comparison sits in transparent pricing.
Frequently asked questions
Sequence against your constraint rather than the market ranking. If your AI initiative is stalling, the answer is almost always data engineering rather than another AI hire. If you are passing security reviews to close deals, it is security. The most common sequencing error is hiring the exciting role before the enabling one.
The 2026 band runs from $118,500 to $190,750, which is the widest on the list, because seniority in security translates directly into risk reduction that boards can see. Projected growth of 29% to 35% against persistent shortage pushes the top of the band up faster than the midpoint. Offering at the midpoint for a senior security hire generally fails.
For most companies starting out, yes. AI programmes commonly stall on data quality, pipeline reliability, and lineage that nobody built, and no amount of model work compensates for that. Data engineering is cited by 22% of leaders as a priority partly because so many organisations discovered this the expensive way.
Specify capabilities instead of titles and years. Name the tools, the scale, and the kind of problem, so that a DevOps posting makes clear whether you need Terraform, Kubernetes, and observability work or pipeline scripting. Vague postings produce large pools with low relevance, which costs you in screening time rather than saving you in reach.
What to do with this list
The ranking tells you where the market is competitive. It does not tell you what you need.
The more useful exercise is working backwards from your actual constraint. What is currently stopping you shipping, passing audits, or controlling costs? That names your role, and it frequently is not the one highest on the list.
Then specify the capability rather than the title, because the variance inside each of these categories is wider than the variance between them.
And accept that for the scarce end of any of these categories, access is the binding constraint rather than budget. To work through which of these roles your situation actually calls for, get in touch.
