The blog
Engineering notes from a productized firm
Plain-English writing on AI-paired delivery, SaaS rebuild economics, regulated-industry architecture, and the operating model behind productized engineering. 51 posts. New writing roughly every 2–3 weeks.
The 2026 software line item: budgeting for the AI premium
AI heads toward half of IT spend by 2027. Split the AI you own from the AI you rent, hold the rented premium flat, and fund a rebuild from the savings.
Read article→
DPDP: 70 days to the November 2026 inflection
Around 13 November 2026 the DPDP Consent Manager framework starts and soft enforcement ends. What changes, and the 10-week build plan to be ready.
Read article→
Agentforce at $2 a conversation: the break-even math
Salesforce priced Agentforce at $2 a conversation and $0.10 an action. Where a custom agent on your own model contract wins the math, and where it does not.
Read article→
Forced AI bundling: the 2026 renewal tax
The 2026 renewal quote has a mandatory AI tier, 20 to 37 percent more. Un-bundle it to save 20 to 35 percent, and rebuild the slice you actually use.
Read article→
The AI-generated app is free. The maintenance bill isn't.
AI generates a working app in an afternoon. The maintenance bill arrives in year two, four times higher. When to vibe-code, and when to engineer it.
Read article→
SaaS renewals in H2 2026: what CFOs are actually cutting
A pattern-match from 40+ H1 2026 renewal cycles: which SaaS lines are being cut, which are being renegotiated, and which are being rebuilt. Concrete numbers, not vibes.
Read article→
The 2026 legacy modernization inflection: why ColdFusion, .NET Framework, and Oracle Forms migrations finally make math sense
Three forces met in H1 2026: AI-paired migration throughput, contractor scarcity crossing critical threshold, license inflation. The rebuild that was infeasible in 2022 is now the cheapest option in your budget.
Read article→
What Claude Opus 4.8 changed for internal build-vs-buy decisions
Anthropic shipped Opus 4.8 in Q2. Three things now cheaper to build than buy that weren’t six months ago — and one that still isn’t.
Read article→
The AI pilot graveyard: why 95% of enterprise GenAI stalls before production
MIT says 95% of enterprise GenAI pilots never reach production. The 5% that do share three architectural choices made before the pilot started.
Read article→
Datadog bill shock: the three signals that justify a rebuild
Most Datadog complaints aren’t rebuild candidates. Three diagnostics separate the bills you fight from the bills you rebuild around.
Read article→
DPDP at T-minus 11 months: the Indian B2B SaaS compliance checklist
Eleven months to the May 13, 2027 DPDP enforcement window. A practical checklist for B2B SaaS leadership teams that haven’t started.
Read article→
The observability cost trap: why Datadog bills spike when you add LLM workloads
AI workloads generate 10–50× more telemetry per dollar of business value. Your Datadog contract didn’t model that.
Read article→
Salesforce at $1.2K/seat/month: the 90-day sunset plan for mid-market
When the CFO asks the Salesforce question — and they will, at the next renewal — here’s the 90-day plan we run with mid-market clients.
Read article→
The real reason your Salesforce renewal was 22% higher
The 22% uplift isn’t a market move. It’s structural. Three procurement plays that work, ranked by leverage.
Read article→
Founder’s guide: how to hire an MVP agency without getting burned
Half of first-time founders we talk to have been burned by a previous MVP build. The pattern is consistent. So is the fix.
Read article→
HIPAA software in 2026: what it costs to build it right
Average US healthcare breach cost in 2025: $10.22M. HIPAA-compliant MVPs cost $30K–$100K to build right. The math is unambiguous.
Read article→
Build vs buy for AI-capable internal tools: a 2026 decision framework
The AI-capable internal tooling space tripled in 2025. Most of those tools are wrong for most companies. Three diagnostics that change the answer.
Read article→
Why your AI cost per query doubled this quarter
AI cost per query is volatile. The roadmap that works in 2026 treats model selection as a separate concern from product design.
Read article→
Harvey AI alternatives: when DIY beats enterprise legal AI
A practical comparison of Harvey AI vs a custom legal-AI rebuild, with the math behind the decision.
Read article→
UAE digital transformation: where to start
UAE buyers want speed, not slideware. The opening engagement that consistently gets traction.
Read article→
AI agents in regulated industries: HIPAA / MAS / DPDP guardrails
Practical guardrails for shipping AI agents into HIPAA, MAS, DPDP, and FCA-regulated workflows.
Read article→
US enterprise procurement: what mid-market sellers miss
The five questions in every US enterprise procurement playbook that mid-market vendors fail.
Read article→
New Zealand SaaS: market entry from the APAC playbook
How NZ SaaS companies use APAC as the proving ground before US market entry.
Read article→
B2B SaaS scaling: $1M to $10M ARR engineering choices
The architecture choices that separate companies that ship the next 9× from companies that stall.
Read article→
Stack rationalisation: an executive brief
A short, board-ready brief on how to rationalise a sprawling SaaS stack without breaking workflows.
Read article→
Mobile-first MVPs: a fixed-price playbook
How to ship a credible mobile-first MVP in 12 weeks at a fixed price you can defend to a board.
Read article→
Salesforce sunset: when to renegotiate vs rebuild
A decision framework for when the Salesforce renewal conversation should become a rebuild conversation.
Read article→
Agent observability: instrumenting LLM workflows in production
When an LLM is your workflow, traces, evals, and rollback are non-negotiable.
Read article→
AI Pod for non-AI organisations: a 12-week reset
How traditional businesses get production AI without rebuilding the org chart.
Read article→
Healthcare HIPAA platforms: an architecture brief
PHI segregation, audit logging, BAA-in-the-loop processors. A 2026 architecture brief.
Read article→
K-12 ed-tech: multi-tenant analytics that actually scales
Per-district isolation, COPPA / FERPA / DPDP overlays, and the data model that survives Year 3.
Read article→
Designing Compliant Data Products: A Strategic Guide for Business Leaders
Privacy-first design as competitive advantage — 20-40% conversion lift, higher exit multiples.
Read article→
ASX-aligned D2C brands: the compliance & cost story
Privacy Act 1988, Consumer Data Right, and the cost discipline ASX-listed D2C brands now operate under.
Read article→
DPDP 101: What Indian Businesses Must Know in 2026
Comprehensive guide to India's DPDP Act — penalties, compliance principles, and a 90-day action plan.
Read article→
MAS-aligned fintech in Singapore: the architecture conversation
Outsourcing guidelines, TRMG, and the architecture decisions a Singapore wealth-tech CTO must make.
Read article→
UK enterprise procurement after Brexit: what changed
Data residency, GDPR-UK divergence, and the procurement questions UK enterprise buyers now ask.
Read article→
Sharing Data With Vendors Via APIs Under DPDP
Defense-in-depth API security architecture for DPDP-compliant vendor integrations.
Read article→
Agents in production: when LLM workflows replace integrations
When the integration cost exceeds the workflow value, agents become the cheaper option.
Read article→
FMCG D2C: the analytics rebuild that pays in 6 months
Klaviyo + Mixpanel + Segment + Shopify Plus add up fast. The rebuild ROI is now under 12 months.
Read article→
Build vs buy in 2026: the new economics
AI-paired delivery has changed the build-vs-buy math. Here's how to think about it now.
Read article→
Run a stack audit before you refactor anything
Refactoring without an audit is treatment without diagnosis. The audit pays for itself.
Read article→
The hidden cost of stack sprawl: a Datadog rebuild walkthrough
When observability tooling costs more than the systems it monitors, it's time to rebuild.
Read article→
Replacing Harvey AI: an AmLaw 200 case file
$686K reclaimed in 11 weeks. Same workflow at <5% of the cost on direct AWS / Anthropic billing.
Read article→
Per-seat SaaS math is broken: when pricing stops scaling
When you're paying $1.2K/seat/mo for a tool 15–25% of your team uses, the math has stopped working.
Read article→
AI Pod vs hiring an ML team: a CTO decision tree
When to embed an AI Pod, when to hire FTEs, when to do both. A practical framework.
Read article→
How AI-paired delivery delivers 3–5× throughput on well-defined work
Senior engineers paired with AI ship faster on well-scoped problems. Here's the pattern.
Read article→
Five outcome-bundled service lines: how we organise engineering work
Build / Rebuild / AI Pod / Agents / Strategy. Each is a container, not a task list.
Read article→
Your code, your repo, your cloud — from day one
IP ownership shouldn't be a hostage negotiation at the end of an engagement.
Read article→
When fixed-price beats time-and-materials
T&M optimises for the vendor. Fixed-price aligns vendor and client around the same outcome.
Read article→
The productized engineering firm: an operating model
Five outcome-bundled service lines. Fixed scope, fixed price, senior-engineer-led delivery.
Read article→
Why we don't sell hours anymore
Selling hours misaligns engineering and outcomes. Here's the operating model we use instead.
Read article→