The decision has been made on evidence - ideally through a Blueprint or an equivalent internal case with five-year math.
Software Insourcing
EXECUTION
One system. Twelve weeks. You own it at the end.
A 90-Day Insourcing Sprint executes one ownership decision in twelve weeks: replacing a specific subscription with an internally owned system, or moving a stable workload off public cloud - delivered with documentation, monitoring, and a named internal owner trained to run it.
USE THIS WHEN
When this service is the right fit.
Use this service when these conditions are present. If the first move is still unclear, start with the Insourcing Readiness Score.
The workload is stable and well understood; the sprint is not for domains still in flux.
A named internal owner exists and has real time allocated - we build with your team, not instead of it.
Leadership accepts that the subscription you delete gets partially replaced by a capability you must run.
WHAT YOU GET
What your team can use immediately.
Each engagement leaves evidence, a decision, and a plan owners can execute — with the independence that flat-fee pricing protects.
Deliverables
- The working system: an internal replacement in production, or the workload migrated and stable.
- A cutover plan executed without business interruption, with rollback tested before it is needed.
- Runbook, monitoring, alerting, and an operational checklist your team actually uses.
- Security and compliance ownership handoff, documented against your obligations.
- The internal owner ramped: paired through the build, not handed a repository at the end.
- A 90-day-after support window and a maintenance plan priced honestly for years two through five.
Where we hold the line
- We do not build in categories that fail the ownership screen - payroll processing, core accounting ledgers, and compliance-heavy systems stay with vendors unless you carry the compliance ownership.
- No sprint starts without a named internal owner; systems without owners become the failure stories.
- AI-assisted development speeds real work, but its savings enter the plan only where demonstrated on your pilot - never as an assumption.
SAMPLE SITUATIONS
The decision belongs in a workflow, not a pitch.
These examples show the before and after state. The actual engagement is scoped around your contracts, systems, spend, and team.
The subscription replacement
- Before
- A workflow tool costs $180K/year for features the team describes as "a form and three reports."
- After
- An internal system does what the business actually needs, with no per-seat fee and the code owned outright.
The cloud exit
- Before
- A stable workload runs on public cloud at a cost that grows faster than the business.
- After
- The workload runs on owned or colocated hardware with monitoring, runbooks, and a real refresh budget.
The 80% rescue
- Before
- An internal build stalled at almost-done and nobody trusts it in production.
- After
- The last 20% - hardening, monitoring, documentation, cutover - is finished and the system carries real load.
HOW WE WORK
Evidence first. Decision second. Independence always.
The cadence is deliberately practical: scope, gather the evidence, run the math, and hand back a decision your team can defend.
- 01
Weeks 1-2: design, cutover plan, and the maintenance model for years two through five - priced before the build starts.
- 02
Weeks 3-8: build or migrate in vertical slices, with your owner pairing on the work from the first week.
- 03
Weeks 9-11: hardening, monitoring, documentation, and a rollback-tested cutover rehearsal.
- 04
Week 12: cutover, stabilization, and the operational handoff - followed by the 90-day support window.
RELATED SERVICES
Choose the next relevant step.
Use these services to move from one decision to the next: audit a contract, triage the stack, execute a build, or install a standing discipline.
RELATED INTELLIGENCE
Operating analysis for practical spend decisions.
These articles cover technical debt, migration risk, unit economics, and financial infrastructure in more depth.
A "perfect" 12-month blended CAC payback often hides a starved enterprise pipeline. Here's the cohort math buyers actually underwrite — and the 88% NRR it exposes.
A SaaS target's 82% gross margin can hide a single-vendor API bill that quietly halves it. How to diligence AI dependency, model drift, and COGS before LOI.
Acquired ML engineers don't quit over money. They quit the week IT puts a two-week ticket between them and a GPU. Here's the post-close playbook that keeps them.
AI targets don't fail in the codebase—they fail in the retraining pipeline. A buyer's field guide to auditing MLOps maturity, model drift, and registry gaps.
A PE diligence playbook for tech M&A: separate a real GenAI moat from a $25/month API wrapper, audit the IP chain, and price inference cost before you sign.
A two-line change to a reporting page shouldn't crash your payment gateway. When it can, buyers cut the price. Here's how brittleness becomes a 22% discount.
FAQ
Questions leaders usually ask.
Why not just hire a development shop?
A development shop is paid to build; whether building was the right call is not its question, and the years-after maintenance cost is rarely in its quote. The sprint starts from a triage that had no build upside, prices the ownership cost before writing code, and is designed around the handoff - the part that determines whether this works in year three.
What happens when the sprint ends - who maintains the system?
Your named owner, ramped during the build, with a runbook and monitoring in place. Companies that want continuity keep us on Managed Internal Platform Support, which exists precisely because the maintenance question is where insourcing succeeds or fails.
Will this actually save us money?
Sometimes - and we will show you the five-year math before you commit, including the case where it does not. The honest record on insourcing is that control, speed, and fit materialize more reliably than headline cost savings. If the case only works on cost savings, we will tell you not to do it.
Can you migrate us off AWS the way 37signals did?
If your workload passes the same preconditions that made their exit work - stable, well-understood workloads and a real operations capability - yes, and their published numbers show what is possible at roughly 75 employees. If your workloads are volatile or your ops bench is thin, the better answer is often optimization or a partial move, and the scoping call will say so.