What investors are funding, and on what evidence

The financing environment is exogenous. It sets the bar regardless of what the company believes about itself.

Three things follow.

The evidence is short of the revenue bar by $1.3M against the general software median of $2.5M, or by $2.3M against the $3.5M the AI-specific reading asks for. An AI company should expect the second. Neither is an order of magnitude, which is the useful point.

Whether the round size itself is a problem cannot be settled from the inputs, because they never say which stage this is. Against a Series A round range of $5-15M, $5-10M is unremarkable. Against a 2026 median seed of $3.1M, or $4.6M for AI companies, it is a large seed. The two readings are judged differently: a seed is priced on evidence of a working motion rather than on revenue, which would make the pilot ladder and the conversion data the things that matter and the revenue gap much less important. Settling which round this is changes which of the two paragraphs above applies.

The margin question is a funding question rather than a presentation question. AI companies are no longer held to the 70-80% once expected of pure software, because inference costs land in cost of revenue, and they are observed to average about 52% in 2026, with the expectation stated as a direction rather than a threshold: a margin trending above 60%, and a company at 55% with a credible path to 65-70% reading better than one claiming an inflated 75% that will not survive scale.

On the comparable basis, the prices Innovera actually charges produce 41.5% at the bottom of its recent band, 56.2% at the midpoint and 64.9% at the top, and 65.5% at the midpoint if the advisor is paid at the low end of the range for the role. The $25K list price at 29.9% is the sharpest commercial problem in the memo.

Whether that band is good or bad depends on a comparison this analysis cannot make. The AI benchmark was relaxed because inference lands in cost of revenue, and inference is 1.4% of what delivering one engagement costs here; the right comparator for a business whose cost of revenue is senior people is a services one, and none is in this analysis. What the numbers support is the range and its direction, not a verdict on whether it clears.

What the bar actually asks for is the one thing nobody has produced: evidence that the number is moving in the right direction.

One source is used for both AI-specific figures here, deliberately. Taking the favorable margin benchmark from the AI-company literature and the favorable revenue benchmark from the general software literature would produce a flattering comparison that no single investor holds.

Revenue per employee is now explicitly used as an efficiency signal for AI companies. The deck sets Innovera beside Palantir on three measures and gives two sets of figures: $1B of revenue at 100 people with a market value above $20B, and $4.4B at 4,400 people with a market value of $300B. The second set matches Palantir's actuals closely (revenue $4,475M, headcount 3,936, market value near $300B), so the first is Innovera's aspiration rather than a second company. The slide does not label which column is which, and that reading is an inference; on any other reading the figures do not correspond to Palantir at any point in its history. Taken as the aspiration, $1B at 100 people is $10.0M of revenue per head, 8.8 times Palantir's actual $1.14M. On its own plan, Innovera reaches $662K of revenue per advisor by 2029 before counting engineering, sales or overhead, which is a services ratio.

The ambiguity is worth removing from the slide. A reader who resolves it the other way concludes the deck has Palantir's numbers wrong.