Rent the ledger, build the screen: where AI actually moved the buy-versus-build line

Andrew Siemer, founder and chief executive of Inventive Group, used a Forbes Tech Council column to argue that the loudest version of the buy-versus-build conversation, in which cheap AI-assisted development puts SaaS in trouble and every company starts building, is also the least useful one ^[1][2]. That matters for anyone holding a renewal calendar, because his case is not that the line collapsed but that it moved: AI changed software economics unevenly, and did not erase the advantages of vendor scale, compliance programs or integrated platforms ^[3].

The part that got cheaper, on Siemer's account, is the layer around the business: the dashboards, internal tools, workflow interfaces and decision-support systems teams live in every day ^[4]. The core system underneath those workflows is still often worth buying; the workflow wrapped around it increasingly is not ^[5]. His reason is unsentimental about what a system of record actually is. It holds the historical data, sits inside a dense integration web, supports audit requirements and serves as the operational source of truth for several teams at once, which makes replacing it expensive, risky and often unnecessary ^[8]. The friction sits elsewhere, in the extra screens, awkward handoffs, narrow tasks buried inside broad software and workflows that almost fit but never quite do ^[9].

The older logic was simpler. Buy-versus-build was mostly settled by cost, vendors spread development spend across many customers, and buyers accepted awkward workflows as the price of avoiding a custom build ^[6]. Siemer's point is that AI did not change the desirability of fit, only the price of getting closer to it ^[7]. Note the evidentiary standard here: he says the cost of building at the edges has dropped enough that old assumptions deserve a second look, and offers no figure for how far [c7b]. His firm sells software product work, which is worth holding in view when reading a piece that argues for building at the edges ^[1].

Two external markers are cited. The Wall Street Journal reported in April 2026 that large companies were not ripping out platforms like Salesforce, SAP and Workday, but negotiating harder, customizing more and building smaller AI-driven tools around them ^[10]. A 2026 paper on agentic AI and enterprise software economics argues the collapse thesis is overstated while finding that building is most compelling for commodity utilities and differentiating custom applications, and much less compelling for regulated and mission-critical systems ^[14]. Both land on the same boundary: the governed core stays rented, the surface is contestable ^[1].

The consequence for budgets is the reframing of breadth. Where a team uses a narrow slice of a large product and the rest mostly adds complexity, the buyer is subsidizing breadth it does not need in order to reach the part that matters ^[11]. Siemer stops short of saying every narrow use case should become an internal build, but he does argue feature breadth is getting harder to defend as an automatic advantage ^[12]. His buy test is narrow and usable: source of truth, real regulatory burden, or a seat at the centre of a dense and valuable integration ecosystem ^[13].

Watch what happens on the second renewal after a team ships its own workflow layer. If the vendor holds price while the used surface shrinks, the tax is real and visible; if vendors start pricing narrow slices, the moat is being rebuilt on commercial terms rather than feature count.