Real rebuilds.
Real reclaim.
3 customer engagements, anonymized for privacy but with full Y1 reclaim math, build cost, vendor stack, and payback period. The kind of detail you can drop into a CFO memo.
Series B B2B SaaS
~120 employees · ~$22M ARR · post-Series B
HubSpot Marketing Pro
Intercom Advanced
Sentry Business
Mixpanel Growth
Y1 net reclaim · payback 4 months
The situation: A B2B SaaS company in product-led growth mode was paying $218K/yr across HubSpot Marketing (sales emails + lead scoring), Intercom (in-app chat + Fin AI), Sentry.io (error tracking, ~30 services), and Mixpanel (product analytics). Their Series B raise put a spotlight on burn rate — SaaS was the third-biggest line item after payroll and AWS.
What we built: A unified customer-and-product platform on Postgres + AWS SES + Anthropic for AI. Marketing emails routed through SES, lead scoring as a Postgres function, in-app chat as a custom widget backed by Anthropic Claude for AI replies, error tracking on self-hosted Sentry (sentry.io's open-source code), product analytics on PostHog. All four wrappers gone, all data in one place.
Build economics: $68K fixed price, 10 weeks, code in their GitHub from day one. Their senior engineer paired with our team for 2 days/week of design review — ensured alignment on data model and integrations.
Y1 math: SaaS spend $218K → direct vendor cost $26K + Allied build $68K + hosting $12K = $106K total Y1. $112K saved Y1 after eating the build cost. From Y2: ~$180K/yr saved. 3-year reclaim crossed $510K.
Quote from CFO: "The build paid back inside Q2. By Q4 we'd shifted the saved budget into a marketing pod. Best swap we've made."
AmLaw 200 firm
~50 attorneys · corporate / litigation practice
Harvey AI
Spellbook (drafting)
+ misc legal-AI tools
Y1 net reclaim · payback 2 weeks
The situation: AmLaw 200 firm with 50 attorneys was paying ~$1,200/seat/mo for Harvey AI plus add-on tools — $720K/yr total. Their managing partner flagged the cost during partnership review. The work being done was primarily contract review, citation handling, and drafting assistance — all backed by GPT-4 and Claude.
What we built: A custom legal-workflow application on Anthropic Claude direct + OpenAI direct. Same matter-context model as Harvey (matter-attorney privilege boundaries, document upload, citation engine). Built-in zero-retention contract with both Anthropic and OpenAI. Documents never leave the firm's tenancy. Code in their GitHub.
Build economics: $24K fixed price, 5 weeks. Heavy security review (firm's IT team and managing partner reviewed every architectural choice). Senior engineer + partner on the engagement.
Y1 math: Harvey spend $720K → Anthropic + OpenAI direct $6K + Allied build $24K + hosting $3.6K = $33.6K total Y1. $686K saved Y1. Build paid back in 2 weeks of saved Harvey billing.
Quote from managing partner: "The cost was so out of proportion that the IT team was certain there was a catch. There wasn't. We got the same workflow at <5% of the cost."
D2C brand
~$80M GMV · ~500K active customers
Klaviyo (email + SMS)
Cloudinary (image CDN)
Algolia (product search)
Yotpo (reviews + loyalty)
Y1 net reclaim · payback 12 months
The situation: D2C brand on Shopify Plus, paying ~$112K/yr across Klaviyo (email + SMS), Cloudinary (image CDN + transforms), Algolia (search), and Yotpo (reviews + loyalty). All four were charging per-contact / per-event / per-record fees that scaled with growth — meaning the SaaS line was outpacing GMV growth.
What we built: A unified D2C customer-data platform. Email + SMS routed through AWS SES + Postmark + Twilio direct. Image CDN as Cloudflare R2 + custom on-the-fly transforms. Product search on self-hosted Meilisearch. Reviews + loyalty as a custom widget on Postgres. All four wrappers replaced as one app, sharing the same customer/order data layer with their Shopify Plus.
Build economics: $58K fixed price, 8 weeks, code in their GitHub. Customer's lead engineer paired weekly. Pilot ran for 3 weeks parallel with the SaaS tools before cutover.
Y1 math: SaaS spend $112K → direct vendor cost (SES, Postmark, Twilio, Meilisearch hosting, R2) ~$8.5K + Allied build $58K + hosting $8K = $74.5K total Y1. $48K saved Y1, then $96K/yr Y2 onwards. 3-year reclaim ~$240K. Plus the data is now unified for future analytics.
Quote from CTO: "Modest Y1 reclaim, but the data unification was the bigger win. We can now do retention analysis we couldn't do across four siloed vendors."
Why these numbers work: AI-paired senior engineers raise the floor on quality + consistency · 250+ products shipped · direct vendor billing — you keep the wrapper margin. → The full method
Want a similar number for your stack?
A 30-min Stack Review delivers a 1-page CFO memo with indicative scope, build cost, and Y1 reclaim — using your actual contracts.