Inference is outrunning revenue
Fine in beta. Brutal at scale.
Home / Industries / SaaS
SaaS
Ship the AI feature without the cost curve behind it.
AI features built so gross margin survives adoption. Sized for your ten-thousandth user, not the demo.
- Cost per user modelled first
- Per-tenant isolation
- Built inside your repo
The starting point
What we hear in the first meeting.
If none of these sound like you, we are probably not the right call yet — and we will say so.
Enterprise buyers are blocking the deal
“A third-party API” is the wrong answer.
Two sprints away for six months
The demo worked. Reliability is a different problem.
Our approach
The demo is the easy twenty percent.
Making it reliable for every tenant, debuggable by support and forecastable by finance is where roadmaps stall. That is the part we do.
How we work →-
Predictable unit economics.
Cost per active user modelled before the build.
-
Isolation that survives review.
Per-tenant boundaries and the documents buyers ask for.
-
Self-hosted where it pays.
We tell you where your crossover point is.
-
Built into your stack.
Your repo, your CI, your observability.
Proof
How we have solved it in SaaS.
Challenge, solution, outcome. Names withheld, numbers not inflated.
B2B platform
- Challenge
- Inference cost grew faster than ARR through the first thousand accounts.
- Solution
- Serving moved to right-sized dedicated capacity with batching and caching.
- Outcome
- Cost per active user flattened, and responses got faster.
Vertical SaaS
- Challenge
- Two enterprise deals stalled on where customer data would go.
- Solution
- Per-tenant retrieval scopes, isolated storage and a documented data-flow pack.
- Outcome
- Security review cleared without an architecture change.
Developer tool
- Challenge
- Every model upgrade shipped on instinct and got reverted on complaints.
- Solution
- An evaluation harness in CI with behavioural regression tests.
- Outcome
- Upgrades became a reviewable diff instead of a gamble.
Teams we work with
Where we plug in
Different teams, different first project. The platform underneath is the same.
AI feature teams
Prototype to a feature every tenant can switch on.
Platform teams
Serving, autoscaling, caching and the cost dashboard.
Security and compliance
Architecture answers for SOC 2 and ISO 27001.
Product leadership
An honest read on what is worth funding this year.
Scope
What we deliver
Four things, in this order. Each one is useful on its own.
AI Discovery
Two weeks to rank the roadmap by value and true running cost.
Feature build
Retrieval, agents or fine-tuning, inside your release process.
Serving infrastructure
Dedicated capacity with autoscaling. Cost tracks usage.
Evaluation harness
Regression tests, so the next upgrade is a decision.
Engagement
How the engagement runs
No surprises about sequence, and no invoice before there is something to look at.
-
Roadmap and cost review
A real running cost per item
-
Architecture
Serving, tenancy and the ceiling
-
Build in your stack
Your repo, your sprint cadence
-
Hand over or run
Your team, or ours under SLA
Under the hood
What it is built on.
Your existing systems stay. We add the layers that are missing and run them.
Your stack
- Your repository and CI
- Postgres and pgvector
- Kubernetes or ECS
- Your observability tooling
AI layer
- Open-weight and hosted models
- Retrieval and agent orchestration
- Per-tenant retrieval scopes
- Evaluation harness in CI
Infrastructure
- Dedicated GPU serving
- Autoscaling and batching
- Response and embedding caches
- Per-tenant cost telemetry
Private cloud
Capacity that scales with your gross margin.
Dedicated serving capacity sized to your real traffic, so inference cost stops tracking signups one for one — and your ten-thousandth user is cheaper than your first.
Explore private cloud- Right-sized GPU serving
- Per-tenant isolation
- Autoscaling and batching
- Predictable monthly cost
What changes
- Flat
- Cost per user as you grow
- 100%
- Tenant-isolated paths
- 4–8 wk
- Prototype to production
Worked examples
The write-ups behind this work.
Full situation, architecture and results on each one.
Legacy modernisation
Changing a system nobody was willing to touch
11,400 Cases pinning behaviour
Cybersecurity
A SOC drowning in alerts that were almost all noise
11k → 40 Reaching an analyst weekly
Infrastructure modernisation
Modernising an estate nobody had a complete map of
400 Hosts mapped, none by hand
Free, no call required
The SaaS AI readiness checklist
The questions we work through before recommending anything — data readiness, hosting constraints, review process and the running cost at year two. Use it with any vendor, including the ones that are not us.
Get the checklist- Where your API-to-self-host crossover sits
- How tenant boundaries are enforced, not assumed
- What a model upgrade has to pass
- Cost per active user at ten times today
Questions
Asked first, every time
Short answers. Longer ones are a conversation.
Should we self-host a model or keep using an API?
It depends on volume. Below a certain steady throughput an API is cheaper and simpler; above it, dedicated capacity wins — often substantially. We run the numbers on your traffic and show you the crossover rather than pushing a preference.
Will you work inside our codebase?
Yes, that is the default. Our engineers work in your repository, branching model and review process. You own everything we write.
How do you handle multi-tenancy?
Data boundaries are designed before anything is built: per-tenant retrieval scopes, isolated storage, and controls preventing one tenant's content reaching another's context. Enterprise buyers probe this hardest, so it gets documented properly.
Can you help us answer security questionnaires?
We provide the architecture documentation, data-flow diagrams and control descriptions those questionnaires ask for. Your team owns the response but is not writing it from scratch.
What if we already have a prototype?
Good — that shortens discovery. We assess what is there, say what carries over to production and what needs rebuilding, then price from that.
Vocabulary
Terms worth knowing before the first call.
The words that come up most in SaaS conversations, in plain English.
Tell us what the roadmap promised.
Send the feature and rough traffic numbers. We come back with an approach and a running cost.
- A reply within one working day, from an engineer rather than an account manager.
- An honest read on whether this is worth doing now, or in a year.
- No marketing list. Your details reach the solutions team and stop there.
Keep looking