Stop Hiring Platform Engineers. Hire One Less.
Most mid-market IT teams in 2026 are losing the same hire. The job description is some variant of “AI Engineer,” “Platform Engineer,” “Automation Engineer,” “ML Ops Engineer.” The loaded cost is $180–280K. The req has been open since Q1. The candidates with real production AI experience are already inside Anthropic, Microsoft, or one of the FAANG-adjacent AI labs. The ones on the open market are 6 months from being productive after they accept.
Here’s the math most CIOs haven’t done out loud.
The cost of the marginal hire
A senior Platform Engineer at $200K loaded ($150K base + benefits + equipment + recruiter fee amortized) costs $16,700/month in cash burn from the day they start. They produce roughly zero net output for the first 8 weeks (onboarding, context, codebase tour, security reviews, first PR landing). They produce maybe 50% of steady-state output for the next 16 weeks (still learning your domain, your customer shape, your incident history, your unwritten conventions). They hit steady-state at month 6.
So the realistic math on a senior Platform Engineer hired in February 2026:
PeriodOutputCostNetMonths 1-2 (onboarding)0%$33K$33K cost, 0 outputMonths 3-6 (ramp)50%$67K$67K for 2 months of equivalent outputMonths 7-12 (steady-state)100%$100K$100K for 6 months of full outputY1 total~8 months equivalent$200K$25K per month of equivalent output
That’s if the hire works out. About 25–35% of senior IC hires at mid-market companies don’t make it past month 9 — wrong cultural fit, better external offer, project misalignment. Account for the 30% failure rate and your effective Y1 cost per month of equivalent output is closer to $36K.
What you’re hiring them to do
Walk through the work the new Platform Engineer is going to take on. At most mid-market IT teams I’ve talked to, the unwritten job description is some mix of:
Add the next 6 vendor formats to the invoice extractor (~2 months)
Wire up the audit logging the GC keeps asking for (~6 weeks)
Build the per-tenant credential scoping the security team flagged (~2 months)
Add SMS as a channel because operations asked twice (~6 weeks)
On-call rotation for the existing pipelines (~ongoing)
The next AI initiative the CIO promised the board (~the rest of the year)
That’s the operational 20% I wrote about in The Last 20%. Items 1-5 are the operational hardening of what your existing team already built. Item 6 is what the engineer thought they were being hired for.
What the engineer wants to do is item 6. What you need them to do is items 1-5. The mismatch is why senior Platform Engineers leave mid-market IT teams after 18 months. They take a senior position at a company doing greenfield work.
The other math: outsource the 20%
Here’s the alternative most CIOs haven’t priced. Instead of hiring the marginal Platform Engineer, contract a specialist whose entire surface area is items 1-5. The going rate at the operations-layer end of the market — for engineering-led mid-market IT customers — is roughly $125K Y1 (engagement fee + first-year support) and $50K/yr steady-state per pipeline.
Compare across 4 years on a single pipeline:
Marginal Platform EngineerOperations-layer specialistY1 cost$200K (load)$125KY2 cost$200K$50KY3 cost$200K$50KY4 cost$200K$50K4-yr total$800K$275K4-yr equivalent output~3.5 yrs (ramp + steady-state - 30% attrition risk)4 yrs (specialist already trained on the shape)Output-adjusted $/yr~$229K/yr~$69K/yr
The specialist is 3.3× cheaper per unit of equivalent output over a 4-year window. They also don’t quit, get poached, or need to be onboarded into the next pipeline.
When the math flips
I want to be honest about when the marginal Platform Engineer is the right hire. Three scenarios.
One. You’re running 5+ pipelines in parallel. At that scale, an in-house specialist is cheaper than 5 separate specialist contracts (per-pipeline ops contracts have natural floor pricing). If you’re past 5 pipelines and projecting 8+, hire the engineer.
Two. The pipelines depend on proprietary domain logic that takes 6+ months to learn. A specialist firm doesn’t have your industry knowledge. If most of the work is encoding your particular regulatory or operational knowledge into the pipeline, the in-house hire makes sense. (Most pipelines aren’t that. Most pipelines are 80% generic with 20% configuration. Be honest about which yours are.)
Three. The strategic narrative requires “we built it ourselves” for board, acquirer, or customer reasons. Sometimes the optics matter. If your customer is a Fortune 100 that audits your tech stack and wants to see in-house engineers on AI, that’s a real reason. (It’s usually a less real reason than the CIO thinks. The Fortune 100 customer mostly wants the audit trail and the SOC 2 attestation, which the specialist can provide.)
If none of these apply, the math is what the math is.
The CFO conversation
The CFO is the unexpected ally on this argument. CFOs at $50M–$1B mid-market companies have spent the last 18 months watching their IT headcount budget grow faster than their revenue. They’re hearing “we need to hire 3 more engineers” at every quarterly review. They have no easy way to push back. The CIO has the technical authority. The CFO doesn’t.
The argument that lands with the CFO is the table above. Loaded cost per unit of equivalent output. 4-year comparison. Output-adjusted dollar per year. CFOs read tables; they don’t read pitches. Bring the table to the next FP&A review and say: “We’re going to hire one less engineer this year and redirect $200K of that budget to a specialist for the operational layer. Here’s the math.”
Your CIO peer at PITT OHIO or any of the other dual-role CFO+CIO companies probably already had this argument internally and arrived at this answer. The single-role CIOs are still hiring.
What I’m actually advocating
I’m not advocating that you stop hiring engineers. Engineers are still the right hire for the 80% of work that’s actual product development, vendor evaluation, domain logic, strategic technology choices. I’m advocating that you hire one less of them and put that $200K-per-year budget into a specialist who runs the operational 20%.
You’ll get more pipelines into production. Your existing engineers will stay longer — they get to work on the parts they want to work on. Your CFO will stop pushing back on every IT req. And your board will see the AI initiative slide actually move from “in pilot” to “in production” — which is the only thing the board really wanted in the first place.
I run one of the specialist firms. That’s my bias. But the math doesn’t care who runs the specialist work — it just says the math.
— Shyan-Ming Perng, Founder & CEO, JieGou — shyan@jiegou.ai