Custom AI-coded apps help small firms trim SaaS expenses

The news: Some small businesses and startups are ditching software-as-a-service (SaaS) giants Salesforce and HubSpot and building custom apps with Anthropic’s Claude Code, Replit, and Lovable instead, per The Information.

TL;DR

A short EMARKETER Briefing (6 July 2026, by Gadjo Sevilla) — a secondary source synthesizing case studies originally reported by The Information and TechGig. Three named cases:

1. Greenleaf Management (55 employees) — replaced Salesforce with an AI-coded app, cutting cost from an implied ~$100K/year to ~$3,600/year. Quote: “Salesforce is super complex… they provide a lot of value, but [we were] only using a little piece of it.”
2. Seattle Seawolves rugby team (70 employees) — replaced both Salesforce and AXS ticketing with bespoke vibe-coded solutions; saved $100K/year in software spend and lifted revenue 25% since March. Anthropic API costs rose, but “the cost benefit is still much better” per team owner Adrian Balfour.
3. Atonom (startup) — cut SaaS costs from $40,000 to $1,200/year after realizing it was paying for enterprise-tier plans sized for much larger teams.

The fork in the road: the article frames this as a strategic choice for enterprise software buyers — deeper SaaS dependency (premium pricing, reliability, continuous vendor-side AI innovation) vs. AI-built custom solutions (subscription fees traded for AI development overhead and vendor/maintenance risk). Explicit caveat: upfront savings may fade if internal teams can’t maintain the custom code or miss future platform updates.

Scope claim (load-bearing): this is framed as a small portion of SaaS providers’ client base — ServiceNow’s cited 97% renewal rate is offered as evidence the core enterprise SaaS business remains largely intact. Implication for brands: full replacement is impractical for most large enterprises (complex custom workflows, data-migration cost, distraction of building proprietary software) — the opportunity is specifically for small companies and those with in-house engineering capacity. Recommended play: test custom agents on specific workflows, then renegotiate SaaS vendor terms as usage decreases, running old and new in parallel to expose gaps.

Why this source matters to the wiki

This is the wiki’s first concrete empirical case-study set for AI-coded SaaS replacement at small-company scale — a sharp, real-numbers counterpoint (not a contradiction) to Sinofsky’s skepticism about vibe-coding into enterprise software. Read together, the two sources draw a scope boundary the wiki didn’t previously have this precisely: small/simple orgs succeed at full replacement; large/complex enterprises (SAP/Salesforce-embedded-logic scale) don’t — Sinofsky’s “the logic and everything else encapsulated in SAP is way more important than the data being in this database” claim explains why Greenleaf/Seawolves/Atonom-scale replacement works (little embedded logic to replicate) while enterprise replacement doesn’t (deep embedded logic).

Linked entities and concepts

• enterprise-ai-adoption — a concrete build-vs-buy data point for SaaS spend.
• dynamic-capabilities — small firms exercising balancing-digital-portfolios at the scale where it’s actually affordable (cf. the corporate-vs-startup asymmetry theme already tracked via McKinsey).
• Anthropic — named beneficiary (token/usage revenue bump from vibe-coding-replaces-SaaS).
• Dangling (single-source mentions, deferred — case-study subjects named in body prose, not author: frontmatter, so the author-entity-promotion rule doesn’t apply): Greenleaf Management, Dave Codrea, Seattle Seawolves, Adrian Balfour, Atonom, ServiceNow, The Information, TechGig.

Source quality

Secondary source — EMARKETER did not conduct the original reporting; the three case studies are attributed to The Information (Greenleaf, Seawolves) and TechGig (Atonom), neither of which the wiki holds directly. Treat the underlying numbers as once-removed — directionally useful but not independently verified by this wiki against the primary reporting. EMARKETER’s own analysis layer (the “fork in the road” framing, the ServiceNow-renewal-rate scope check, the brand-implications advice) is first-party EMARKETER editorial judgment. Short-form “Briefing” format — comprehensive within its scope, not a deep investigation.