FishDog Research studies · Enterprise IT budgets v2.0 · 2026-07-11 · n=10
Shared research study · Fielded May 27, 2026

IT Budgets: SaaS vs AI vs Custom Builds

Ten US enterprise IT budget owners — CIO, IT, engineering and systems managers, including a regulated FDA/ISO voice and a K-12 public-education operator — on how software spend is being reallocated across existing SaaS, AI-native tools, and custom builds. Framed for the software investors and analysts who need to read demand trajectory, switching behaviour, and wallet-share migration.

10enterprise IT decision-makers
70responses · 7 questions
May 27, 2026fielded
US enterpriseincl. regulated + K-12
The market signal this study supports

AI is not a new budget line — it is a reallocation. Watch the leading indicators of wallet migration: AI approaching ~15–20% of the software mix, seat compression in notes / grammar / BI-viewer categories, pilot pass rates, vendor-count decline, and renewal-driven switches fuelled by bundling.

Basis: budgets flat with AI funded by SaaS consolidation (all 10); AI spend cannibalised from existing lines under 60–90 day KPI-gated pilots (9/10).
The contested question · Q3, build vs buy

Ten IT leaders.
Three answers on build vs buy.

The panel converges hard on "hold flat and reallocate" — but splits where the seniority and sector fault lines meet: has AI actually changed the build-versus-buy calculus? Four say consolidate into the suite; four have become build-lite converts; two are frozen by regulation and privacy.

4 consolidate into suite-native AI; building is a maintenance liability 4 build-lite now beats vendor SKUs on unit economics 2 validation & privacy freeze the calculus

Q3 of 7: "Has AI changed the build-versus-buy calculus for your team? Two years ago, building a custom tool meant…" · Position split analyst-classified from the study's segment and divergence analysis — see methodology.

Position spectrum · Q3

Where each leader stands

Hover any respondent to read their position, as reported in the study analysis. Colour follows the three camps above.

◄ Consolidate into the suiteBuild it in-house ►
RN
JC
NA
AE
JM
AM
MH
MM
RM
AMo
Two-letter codes are initials; positions are drawn from the study's per-respondent attributions.
The argument

The two poles of build vs buy

The consolidation case
Suite-native AI with contractual protections, one-in / one-out — top-down consolidation and ROI mandates drive ~60% of the budget motion.
Raymond Navarro · IT budget gatekeeper, Chino CA — with James Cheng (AI 2%→19% via consolidation) and Nicholas Ausbie alongside
VS
The build-lite case
A four-week internal build replaced the vendor SKU at ~60% lower unit cost, 90–95% accuracy, with a small human-review lane.
Aaron Monahan · FinOps CTO, Franklin TN — with Matthew Hughes, Miles Murad and Ryan Maciel running RAG / triage / doc-intake pilots

Both poles share the same discipline: narrow, short-payback, governed. Build-lite wins where the problem is small and maintenance is light; the suite wins where scale, SLA, validation or broad coverage are required. The systems of record — ERP, CRM, HRIS, QMS/ALM — stay with vendors in every account.

Procurement gates

What a new AI purchase has to clear

One row per respondent, one column per condition the study ties them to. Column totals rank the market's demands — every dollar of AI now passes a KPI-gated pilot.

Respondent Data-use
guarantees
KPI-gated pilot
+ ROI offset
Cost predictability
(caps, FinOps)
Suite-native /
consolidation
Validation &
auditability
Ryan Macieltech PM · San Jose
Raymond Navarrogatekeeper · Chino
Miles Muradops research · Oakland
Jeremy Millerregulated · rural NJ
Abigail McclungK-12 ed · rural NY
James Chengops GM · Boston
Matthew Hughessystems analyst · Austin
Nicholas AusbieCTO · Hollywood FL
Aaron MonahanCTO · Franklin TN
Apryl Ellisonsystems analyst · rural MO
demanded by 6/1010/104/107/107/10
  tied to the condition in the study   consistent with their segment (inferred)   hard block — will veto   not surfaced
Consensus and contest

How the panel divides across all seven questions

Six of the seven questions converge hard. Build-versus-buy is the one real fault line. Splits are analyst-classified from the study's segment and divergence analysis, not from per-response coding.

Q3
Build vs buy — consolidators vs build-lite converts vs constrained sectors
Q6
Procurement change — everyone tightened; intensity diverges (validation paths vs FinOps caps)
Q2
Replacing SaaS with AI — most prune point tools at renewal; regulated + education won't swap
Q1
Budget shift — flat budgets, AI to mid-teens via reallocation; K-12 the lone structural outlier
Q4
Additive vs cannibalising — almost entirely cannibalised; one reports partial fresh security spend
Q5
Vulnerable categories — near-total agreement: notes, grammar, BI viewers, forms, async video compress first
Q7
CFO narrative — "hold flat and reallocate," outcome-backed only; education runs a grant-driven logic
The panel

Who answered

10 respondents recruited from a census-grounded synthetic population of 340,000 US residents, with a regulated FDA/ISO voice and a K-12 operator recruited deliberately to capture conservative outliers.

Respondents
10
enterprise IT decision-makers · 7 questions · 70 responses
Fielded
May 27
2026
Median income
$158K
range $76K–$405K · one regulated-sector outlier
Sectors
4
commercial tech · regulated eng · K-12 education · FinOps-led

Panel income vs. US households

Benchmark: US Census Bureau, 2022 ACS 1-year (Table B19001).
Under $50K
0%
35%
$50K – $100K
30%
29%
$100K – $150K
10%
17%
$150K – $200K
30%
9%
$200K+
30%
12%
This panel (n=10) US households

Behind every respondent

Each profile carries a grounded biography and an ingested media diet.
Every respondent carries a grounded biography, an occupation, a region, and the recent news they actually read — this panel's diet runs from NPR Technology and BBC News – Business to local dailies like the Mercury News and Houston Chronicle. The roster below leads with role, region, the build-vs-buy camp, and the position that shaped each answer.
Raymond NavarroChino, CA · 57
Consolidate · IT budget gatekeeper
Attributes ~60% of budget motion to top-down consolidation; buy/build posture shifted ~90/10 → ~80/20.
James ChengBoston, MA · 57
Consolidate · ops GM
AI share up from ~2% to ~19%, funded by suite consolidation and seat reclamation.
Nicholas AusbieHollywood, FL · 57
Consolidate · CTO
Drives procurement clauses; only fresh budget is risk/security under board pressure.
Apryl EllisonRural, MO · 35
Consolidate · systems analyst
Cancelled most Grammarly Business once Microsoft Editor + Copilot sufficed.
Aaron MonahanFranklin, TN · 47
Build-lite · FinOps CTO
4-week internal build replaced a vendor SKU at ~60% unit-cost reduction, 90–95% accuracy.
Matthew HughesAustin, TX · 41
Build-lite · systems analyst
"Allergic to hype"; runs RAG / triage / doc-intake pilots under strict governance.
Miles MuradOakland, CA · 39
Build-lite · ops research analyst
Evidence-over-hype; drives measurable build-lite pilots held to CFO-grade proof.
Ryan MacielSan Jose, CA · 39
Build-lite · tech PM
Quick ROI checks; dropped standalone transcription for suite AI; funding conditional on a quarter's productivity.
Jeremy MillerRural, NJ · 49
Constrained · regulated FDA/ISO
Validation, traceability and provenance freeze vendor swaps; AI augments validated systems only.
Abigail McclungRural, NY · 42
Constrained · K-12 education
SaaS ~93% / AI ~3%; FERPA and offline reliability block public LLMs for student PII.
Executive summary

The one page

FishDogEXECUTIVE SUMMARY · v2.0 · 2026-07-11

Enterprise AI spend is a reallocation, not a new budget line — and build-lite is now real

Ten US enterprise IT budget owners describe flat budgets with AI grown from low-single-digits to mid-teens share, funded by SaaS consolidation and seat right-sizing. The panel converges on "hold flat and reallocate," and splits only on whether AI has changed the build-versus-buy calculus.

  • 4 consolidate, 4 build-lite, 2 are frozen. Suite-native AI wins on governance and coverage; narrow, short-payback internal builds now beat vendor SKUs on unit economics (one replaced a SKU at ~60% lower cost in four weeks). Regulated and K-12 contexts stay bundled and minimal.
  • Cuts hit the edges: meeting transcription, writing/grammar, BI viewers, forms and async video compress first. Systems of record — ERP, CRM, HRIS, QMS — are augmented, never replaced.
  • Every AI dollar passes a gate: 60–90 day KPI-gated pilots (10/10), no-train data guarantees, retention/residency controls, prompt/output logs, and hard spend caps.

Signals to track: AI at ~15–20% of software mix, seat compression in notes/grammar/BI-viewer categories, pilot pass rates, vendor-count decline, and renewal-driven switches.

Synthetic panel study, n=10, fielded 2026-05-27 · full split, procurement matrix and methodology above and at fishdog-report-lab.pages.dev/it-budget · Directional; positions on the build-vs-buy split are analyst-classified.
↓ take the data with it
Methodology

How this study was made — and where to be careful

Respondents
Synthetic. 10 US enterprise IT decision-makers recruited from FishDog's census-grounded population of 340,000 US residents; each carries a grounded biography, occupation, region and media diet. A regulated FDA/ISO voice and a K-12 operator were recruited deliberately as conservative outliers.
Fieldwork
7 open-ended questions, fielded May 27, 2026 · 70 responses (10 per question).
Positions
The study's analysis reports per-respondent evidence as attributed reported-speech (e.g., "James Cheng reports AI rising from ~2% to ~19%") rather than as first-person verbatim quotes, so this page attributes positions rather than quoting them. The Q3 build-vs-buy split (4/4/2) and the per-question divergence bars are analyst-classified from the study's segment, shared-mindset and divergence tables, not from per-response coding — treat the counts as illustrative of the fault line, not as a precise vote.
Read with care
Qualitative, n=10, synthetic — directional, not a market census. Six of seven questions are near-consensus; the value is the build-vs-buy fault line and the procurement gates, both of which map cleanly to wallet-migration signals.
Every answer
The full record ships as data, not a link. All 70 responses, verbatim, with each respondent's biography, are in responses.csv ↓ (respondent × question × verbatim) — open it in any spreadsheet and run your own analysis. Respondent-level stances: stances.csv. Live study: shared link.

FishDog · Research without respondents · Fielded May 27, 2026 · 10 respondents · 7 questions · 70 responses · v2.0, rebuilt 2026-07-11. Positions and demographics are from the study record; the build-vs-buy split is analyst-classified from the study's divergence analysis. How this report is structured.