Orcher · Component 06 of 07
Routes by stakes and halts runaway loops.
Layer 2 — continuous, non-blocking
Not every directive deserves the frontier. Cost Governance routes by the stakes of the action rather than the habits of the developer, enforces budgets at the role and department level, and terminates loops that would otherwise spend without producing an outcome. Cost is a governance property, not a finance report written after the fact.
Without it, spend outruns value
Ungoverned agent fleets exhaust annual budgets in a quarter and produce no auditable outcome to show for it.
What it emits
A routing decision and a cost record per call, reconciled to the directive.
Evidence
Figures drawn from the Enterprise Superintelligence Report, Vol. I, August 2026.
40–85%
spend reduction observed under Orcher cost governance
range across deployments — Vol. I
4,500×
price spread across 400+ models and 70+ providers
Vol. I, p.25
63%
cost reduction at 91% accuracy match in internal routing tests
Dipp AI Research: internal analysis, not independently verified
68%
of enterprise AI programs are over budget
Vol. I, p.9
Mechanism
Four stages, in order. Layer 1 components gate execution; Layer 2 components run continuously and never block.
01
Score the stakes
Each directive is scored by consequence: what changes in a system of record, whose authority is exercised, and what a wrong outcome would cost. Stakes, not developer habit, set the tier.
02
Route to a tier
Across 400+ models and 70+ providers the price spread is roughly 4,500×. Orcher selects from that range deliberately, and escalates only where verification demands the stronger model.
03
Enforce ceilings
Budgets are enforced per role, per department, and per directive class at call time — not reported after the quarter closes.
04
Detect and halt loops
Non-terminating agent loops are detected and stopped, and the halt is written to the ledger like any other outcome rather than disappearing into a bill.
Operational contract
What it is not
It is not a FinOps dashboard
Dashboards describe spend that already happened. Cost Governance decides, at call time, whether the spend is permitted at all.
It is not a race to the cheapest model
High-stakes directives are routed to the strongest available tier on purpose. The discipline is matching the tier to the consequence, in both directions.
Failure behaviour
When a ceiling is reached or a loop is detected, execution halts and the directive is recorded as halted with its cost trace. Spend never continues silently past a limit.
Questions
Where it shows up
Surfaced automatically from the components each solution engages and each industry relies on.
Solution
Route by the stakes of the action, not the habits of the developer.
Solution
Utilization is a governance outcome, not a procurement problem.
Use case
Every adjudication is a decision someone must own.
Use case
Supervised institutions need evidence, not dashboards.
Use case
Margin decisions at machine speed still need an owner.
The strategy behind the component
Binding a platform to one heavy frontier model wastes tokens and leaks proprietary data. Dipp AI's strategy — and the reason Orcher's seven components exist — is intelligent routing: decouple orchestration from any single provider, match each task to the cheapest model that can carry it, and keep prompts and data inside the enterprise's own tenant boundary.
DIPP AI / ORCHER
Seven components. One enterprise-owned control plane.
01 + 02 / AUTHORITY
Intent, named professional, permitted actions and systems of record.
06 / COST GOVERNANCE
Select the least costly sufficient model within the task’s stakes, latency target and spend ceiling.
Explore Cost Governance05 / DATA CONTROL GATEWAY
Region, redaction, retention and training-exclusion requirements determine which destinations qualify.
Models discussed in Dipp AI Research · September 6, 2026
Model families · exact versions and licences assessed per deployment
Hyperscalers, neoclouds and enterprise infrastructure
03 / Logic Scrubber
Verify proposed actions before commit.
04 / Immutable Audit Ledger
Preserve the decision and refusal evidence.
07 / Observability
Trace cost, latency and outcomes across routes.
Switch providers in a week, not a quarter.
An orchestration layer sits in front of every AI task rather than inside one vendor's SDK. Directives, roles, policy and evidence live in Orcher, so a provider change is a routing-table change — a week of work rather than a quarter of re-platforming — and no model vendor becomes the system of record for how the enterprise operates.
The protocol stackRoughly 90% of workloads never need the frontier.
Cost Governance routes by the stakes of the action rather than the habits of the developer. Standard, routine queries — roughly nine in ten workloads — go to cheaper open-weight or small in-house models; expensive frontier models are reserved for genuinely complex, high-reasoning tasks. Across 400-plus models and 70-plus providers the price spread is 4,500× (Vol. I, p.25), so the routing decision is the economics.
Cost GovernanceProviders never learn from your interaction data.
The Data Control Gateway forces data and prompt histories to stay inside the company's own cloud environment — a private tenant, in-region — and verifies training exclusion and the data boundary on every call before a token leaves. Using frontier capability never means donating enterprise interaction data to the firm that sells it back.
Data Control GatewayInteractive · routing simulator
Each workload type carries a different level of stakes, sensitivity and reasoning depth. Select one to see which Orcher component owns the decision, which model class it routes to, which gates run, and what that does to tokens and cost against a frontier-default baseline.
Workload type
Routing decision
High-volume intake: label the document, extract identifiers, route it to a queue. No irreversible write, no judgement call.
Decided by
Cost GovernanceRouted to
In-house small model
Runs inside the tenant. No egress, lowest unit cost, deterministic latency.
Token reduction
42%
1,800 → 1,044 tokens
Cost per directive
$0.00013
baseline $0.0432 on frontier model
Cost saving
100%
2360ms faster median
Gates that run before commit
Stays inside the tenant entirely — the payload never reaches an external provider.
Indicative class prices per 1M tokens: In-house small model $0.12 · Open-weight mid model $0.60 · Commercial mid-tier $4.50 · Frontier model $24.00. Illustrative only; the 4,500× spread across 400+ models and 70+ providers is from The Enterprise Superintelligence Report, Vol. I, p.25.
Interactive · economics
A frontier-default estate pays the top price band for every workload, including the nine in ten that never needed it. Move the inputs to your own volumes and see the token waste removed and the cost impact of the routine-versus-frontier split.
Your estate
One directive is one verified execution cycle.
Prompt plus completion, before scoping and redaction.
Directive scoping, field redaction and cache reuse before a model is called.
Your blended frontier rate.
In-house or open-weight class hosted in your tenant.
Estimated impact
Monthly cost, frontier default
$30.0K
1.25B tokens at the top band
Monthly cost with routing
$2.5K
$459 routine · $2.0K frontier
Token waste removed
400.0M
32% of 1.25B never reaches a model
Cost reduction
92%
$27.5K per month
Annualised
$330.0K
Retained by routing 90% of directives to a routine class and reserving the frontier for the 10% that genuinely needs it — with every routing decision hashed into the same audit chain as the action it paid for.
Estimates only, for planning conversations rather than contract terms. The 4,500× price spread across 400+ models and 70+ providers, and the 68% of AI programmes running over budget, are from The Enterprise Superintelligence Report, Vol. I (pp. 25, 9).
Proof points
Dynamic model routing is only a strategy if it moves numbers a finance team and an auditor both recognise. These are the outcomes it is accountable to.
Cost
4,500×
price spread across 400+ models and 70+ providers
The spread between the cheapest sufficient model and the frontier default is the entire economic case for routing. A directive sent to the wrong class does not fail — it just costs orders of magnitude more than it had to.
Vol. I, p.25
Token waste
~90%
of workloads never need a frontier model
Routine classification, extraction, summarisation and drafting clear on in-house or open-weight classes. Reserving frontier capacity for genuinely complex work is what turns AI spend from a run-rate into a budget.
Dipp AI routing model · Vol. I, p.25
Budget
68%
of enterprise AI programmes run over budget
Overrun is a control failure, not a forecasting failure. Cost Governance applies a ceiling per role and per directive and halts loops at the step boundary rather than at the invoice.
Vol. I, p.9
Latency
10×
faster on routine classes than a frontier default
Cheaper classes are also materially faster. Routing the routine nine-tenths off the frontier shortens the median directive as a side effect of the economics.
Indicative class latencies, Orcher routing table
Auditability
28%
of enterprises can trace an agent action end to end today
Every routing decision Orcher makes — class chosen, alternatives considered, ceiling applied, boundary attested — is hashed into the same evidence chain as the committed action, so the economics are as auditable as the outcome.
Vol. I, p.6
Portability
1 week
to change providers, not a quarter of re-platforming
Because orchestration is decoupled from any vendor SDK, a provider change is a routing-table change. Directives, roles, policy and evidence stay in Orcher.
Dipp AI protocol stack contract
The other six
Layer 1
Plain language becomes the permanent record of what was asked.
Layer 1
Binds the directive to a human and a role, cryptographically and revocably.
Layer 1
Verifies the proposed action against systems of record before commit.
Layer 1
Hashes the verified action permanently. Evidence, not logs.
Layer 2
Training exclusion and data boundaries verified before routing.
Layer 2
Real-time cross-provider trace: which model, which role, what cost, what outcome.
Further reading · September 2026
Two long-form pieces per week through September 2026 — sourced, dated, and attributed. Each one links back to the components it argues about.
Sep 21, 2026
Model routing is not new. Almost every gateway vendor offers it. What is still missing is the difference between a router suggesting the cheaper option and a system that will not let the expensive one run unless the stakes justify it.
By Odero Otieno
Sep 23, 2026
The enterprises winning the next five years will not be the ones who bought the most GPU capacity. They will be the ones whose capacity is actually doing verified work — and one 23,000-cluster study suggests almost none of it currently is.
By Odero Otieno
Sep 14, 2026
The industry has spent 2026 trying to bill AI agents for outcomes instead of tokens. Almost none of it has solved the harder problem underneath: proving an outcome actually happened before anyone gets billed for it.
By Odero Otieno
Sep 9, 2026
Seven components, two layers, one verified execution cycle — built so autonomy can run at machine speed without the enterprise losing the ability to prove, on demand, what happened and under whose authority.
By Odero Otieno
Orcher is deployed with named enterprises under the Human-in-the-Role model. Request a technical briefing with the founding team.
