Where state-of-the-art AI software is landing by industry: impact ranges and time-to-value pulled from our actual builds. Anonymized, structured, published on a quarterly cadence.

Why the index exists

Most of what gets published about AI outcomes is either vendor marketing or survey data, and neither is instrumented. Vendors publish case studies that skip the base rate. Surveys publish sentiment that does not touch the income statement.

The Vertical AI Index is built the other way around. Every data point comes from a real build, with a measured baseline, the software we shipped, and a time-to-value outcome we tracked ourselves. We anonymize at the client level, aggregate at the industry level, and publish a structured quarterly issue.

What the Q3 2026 issue will contain

  • Industry benchmarks. Where state-of-the-art AI software is landing across the industries we build for: professional services (legal, accounting, consulting), multi-location operators, healthcare services, distribution and specialty retail, and founder-led operating companies.
  • Impact ranges. Ranges, not point estimates, with sample size disclosed. We will publish the distribution, not just the median.
  • Time-to-value distributions. How long it actually took from kickoff to measurable P&L impact across build types. This is the number everyone wants and nobody publishes honestly.
  • Failure mode analysis. Where builds stalled, what caused the stall, and how we fixed it. The thesis requires us to be honest about our own misses.
  • Year-over-year deltas. Starting with issue two. What moved. What did not.

How we structure each data point

Every build we run is instrumented the same way, which is what makes aggregation honest. The schema behind every number in the index:

  • Industry and workflow. The vertical and the expertise-bound workflow the software was built for. (See our partners.)
  • Baseline. The dollarized cost of the current state, measured before anything ships.
  • What we shipped. The state-of-the-art software we built, operated, or replaced.
  • Outcome. The measured change on the income statement, with the measurement window and attribution method disclosed.
  • Time-to-value. Days from kickoff to the first measurable P&L signal, and days to stable impact.
  • Confidence. How clean the attribution was. Everything reported honestly, including the weak attributions.

What it will not contain

A few things we will deliberately exclude, because including them would make the index worse:

  • Vendor comparisons. We do not rank tools. We are tool-agnostic by design.
  • Named clients. Every data point is anonymized at the client level. The sector and revenue band appear. The name does not.
  • Predictions. We publish what happened, not what we think will happen. The thesis lives in the writing, not in the benchmark.
  • Survey data. The index is not a survey. If a number is in the index, it came from software we built and operated.

Sample bias, named

We will also publish what the index cannot do. Our sample is the builds that reach us, which skews toward:

  • Operating companies, not venture-funded startups.
  • The mid-market, not the largest enterprises or the smallest micro-businesses.
  • The southeastern U.S., especially the Atlanta metro and Georgia more broadly, though we work nationally.
  • Founder-led firms and professional-services firms, because those are the two groups that self-select into our work.

These biases will be disclosed at the front of every issue. If you want to use the index to calibrate a decision outside our sample, read the disclosure first.

How to receive it

No form. No gate. Email hello@peachstateai.com with "Vertical AI Index" in the subject line and we will add you to the distribution. The Q3 2026 issue will arrive in your inbox on launch day. No other mail.

If you want the quarterly issue but do not want to be on any other list, that is fine. The distribution is single-purpose.