Sources and provenance

Read alongside the confidence-tag definitions in the course index — this lesson doesn't repeat that framework, it applies it to the whole course at once.

---

#How this course was built

This course was researched in a single pass in August 2026, entirely from public web research — no local source corpus or disclosed operator playbook existed for this topic going in, unlike this platform's courses built from one practitioner's own documents. One candidate source was checked directly during this research and specifically declined: an arXiv preprint arguing that agentic AI changes build-vs-buy economics, which turned out to be a non-peer-reviewed, thinly-argued preprint with no clear institutional affiliation and no verifiable data behind its claims. It is named here as a record of that check, not as a citation — the disqualification pass this platform's source-finder methodology requires applies to preprints and papers, not just to named practitioners.

#Named, disclosed-tier sources

• Oliver Williamson's transaction cost economics (2009 Nobel Memorial Prize in Economic Sciences), specifically the asset-specificity concept — the academic foundation for Module 1's switching-cost mechanism.
• AICPA (American Institute of Certified Public Accountants), its own published SOC 2 Trust Services Criteria framework — the standards-body source for Module 2's security-and-compliance lesson.
• Gartner, cited by name for two distinct, dated claims: its widely-reported buying-committee-size research (commonly attributed to its "Future of Sales" work, cited in Module 2, corroborated across multiple independent secondary sources but not independently confirmed against Gartner's own paywalled original) and its June 2025 press release predicting over 40% of agentic AI projects will be canceled by the end of 2027 (cited in Module 4).
• McKinsey Digital and the University of Oxford's BT Centre for Major Programme Management, their joint study of more than 5,400 large IT projects (initial 2012 publication, still the most-cited figure of its kind) — the disclosed-methodology source for Module 3's build-vs-integrate cost data.
• MIT Media Lab's Project NANDA, "The GenAI Divide: State of AI in Business 2025" (Challapally, Pease, Raskar, and Chari; July 2025) — cited in full in Module 4, alongside named published criticism of its methodology, including a specific call from reviewer Kevin Werbach for the underlying data to be released or the finding retracted, rather than repeating its headline figure uncaveated.
• Miller Heiman's Strategic Selling, and CEB/Gartner's later Challenger Sale research — the practitioner-research origin of the economic-buyer/technical-buyer/user-buyer vocabulary used throughout Module 2.

#What's aggregated from cross-checked industry-benchmark sources, not a single disclosed study

Module 1's ACV and sales-cycle-length bands, Module 2's SOC 2 audit-cost ranges, Module 3's enterprise-pricing-negotiation practices, and Module 5's enterprise-AE compensation and startup-funding benchmarks all draw on multiple independent SaaS-operations, compliance-industry, and venture-funding sources that were cross-checked against each other for internal consistency, rather than one named census. Each is tagged [Directional] where it appears, consistent with that sourcing shape.

#What's flagged [Speculative] — the numbers most worth re-checking

• The "150–200% switching-cost exit tax" and "16x higher switching costs" figures (Module 1) — traced to vendor and consultancy content-marketing sources with a direct commercial interest in the claim, no disclosed methodology found; named and set aside in Module 1 rather than repeated.
• The "~30% faster growth from structured discount governance" figure (Module 3) — traces to a single pricing-strategy-firm source without an independently verifiable methodology; the underlying logic is plausible, the specific number is not independently confirmed.
• The exact week/month figures in Module 1's sales-cycle-length table and Module 2's RFP-timeline estimate — the direction (cycle length rises steeply with contract value; a full RFP cycle runs roughly 6–12 weeks) is well corroborated across sources; the specific week counts are benchmark aggregation, not a single authoritative study, and should be treated as illustrative ranges.

#What this course deliberately did not do

Consistent with this platform's practice on courses built from general research rather than a disclosed operator's own numbers: this course does not name or hold up any specific real company, software vendor, or venture as a case study or example to copy. Every mechanism argued in Modules 1 through 4 is sourced to a named research body, standards body, or cross-checked industry benchmark instead, deliberately, so the argument holds regardless of which vendor, platform, or specific market a reader ends up building in.