Every platform ships its own governance pane. Standing Watch is a cross-platform layer that verifies behavior and composes one evidence record across them.
Every major platform vendor now ships its own AI agent governance tool and calls it the layer of record — and each is structurally anchored to its own estate. Only 13% of organizations believe they have the right governance in place to manage AI agents (a Gartner estimate, cited in SAP's August 3, 2026 agent-sprawl article), and 94% of organizations surveyed are concerned about AI sprawl while only 12% have a centralized platform to manage it (OutSystems research, April 2026, nearly 1,900 global IT leaders). Run three or four of these platforms and you own three or four "single panes of glass" — and still have no cross-platform layer that verifies agent behavior and composes one evidence record across them, which is the kind of record a deployer of in-scope high-risk systems, or a NIST/ISO program, will be asked to produce.
Who this is for
CIOs, CAIOs, and compliance leaders running AI agents across more than one enterprise platform — SAP, Workday, Databricks, ServiceNow, or a custom MCP estate — who need one verifiable, auditor-legible evidence story across them, not four competing single panes of glass.
What you get
A cross-platform verification and evidence layer (we review across platforms; our own agent-building work is built on Claude) that sits on top of your existing platform tools and control plane rather than replacing them: an agent qualification register, a cross-vendor spend arbitration baseline, a tiered autonomy policy with a hard human ceiling, behavioral control verification run identically across every system in scope, and a findings ledger that ages instead of silently disappearing.
The estate today
Four vendors, four self-claimed “layers of record”
Each pane below quotes the vendor's own material or on-record statements, as reported — not Tioga's characterization. Standing Watch's six disciplines sit above and across all four, as the cross-cutting layer none of them ship.
Standing Watch — six disciplines, cross-cutting
SAP
AI Agent Hub
“governance layer of record for the enterprise agent ecosystem”
SAP CTO Philipp Herzig, SAP News, Aug 2026
Workday
Agent System of Record
“the single source of truth for all of an enterprise's AI agents”
Workday ASOR GA blog, blog.workday.com
Databricks
Unity AI Gateway
“a unified governance layer for both AI assets and AI interactions”
Databricks blog, Data + AI Summit 2026 (Jun 2026)
ServiceNow
AI Control Tower
“we manage them where technically possible”
ServiceNow product leadership, per Techzine Global reporting, May 2026
A static positioning graphic, not a live data feed — quotes verified 2026-08-10 against each vendor's own material or on-record statements; see source line under each pane.
Why this is real, not a pitch deck
Built and run on Tioga's own cross-platform infrastructure
The six disciplines below are generalized from automations Tioga actually operates in production — the JARVIS router, router-watch, and security-watch — across a genuinely heterogeneous stack (a free local model, multiple OpenRouter-hosted vendors, Google-billed Gemini (free tier first, paid past the limit), and a subscription Claude that's never auto-routed). This is a single-operator, personal-infrastructure-scale implementation, not an enterprise deployment — volunteered here, not extracted in a pitch.
Propose-and-approve by default
Findings are proposals, not changes: nothing is applied on a finding's own authority. A small, hard-coded set of bounded change types (pure additions, syntax-checked afterward) may be applied automatically; everything else becomes a dated proposal a human reviews and applies — a durable, framework-mapped record of agent actions plus human approvals, composed independently of any single platform vendor.
Six disciplines, one running mechanism each
Qualify (capability floors enforced at routing time), Arbitrate (compare spend across non-fungible budgets on one basis), Gate (tiered autonomy with a structurally unreachable top tier), Probe (the same behavioral check run identically across every system in scope), Track (an aging findings ledger — nothing clears silently), Review (fixed-cadence, criteria-based vendor/model reappraisal).
See the real ledger, not a mockup
The Governance Ledger demo is a dated, real excerpt from Tioga's own AI routing gateway, mapped to the NIST AI RMF — the same infrastructure Standing Watch's Track and Review disciplines are generalized from.
Works alongside your control plane
If you already run an AI gateway or agent-security control plane — for discovery, identity, policy enforcement, or a kill switch — keep it. Standing Watch does not build or replace one, and it does not compete on discovery breadth or runtime enforcement: a continuously running product will beat a manual pass on both.
What it adds is the part a single product's pane is least placed to say about itself: cross-platform, fixed-cadence behavioral verification run identically across every system in scope, spend compared on one basis across non-fungible budgets, and one findings ledger and evidence pack mapped to NIST AI RMF, ISO 42001, and the EU AI Act that a named human owner reviews.
Scoping starts with one question: which gateway, control plane, or platform governance tool do you run today? The Assessment uses whatever inventory or log export that tool can produce — a CSV or a report is enough — and says plainly where a control-plane product is the right next purchase.
Engagements
Every engagement starts with a 5-day Discovery Sprint ($5,000 flat, prototype included) that scopes the work before any larger commitment — credited toward the price below if you move forward.
Not ready to scope an engagement? Free ERP Agent-Readiness Checklist →Standing Watch Assessment
Reconciles the agent inventory you already have — from your control plane, platform exports, or a manual list; Tioga does not compete with discovery tooling — and adds what those tools don't: a qualification register and autonomy-tier map, a cross-vendor spend arbitration baseline, a first behavioral probe run across two or more systems, and a seeded findings ledger you keep.
$15–35K
3–4 weeks
Standing Watch Build
Implements the behavioral probe harness and the propose-and-approve review workflow in your environment, on your credentials and repositories, on top of whatever control plane or gateway you already own — modeled on router-watch and security-watch. Tioga does not build a gateway, policy engine, or discovery product, and will not be a required runtime dependency.
$60–150K
8–16 weeks, scoped to estate breadth
Standing Watch Retainer
Router-watch and security-watch as a service, generalized to your estate: a weekly automated watch run (probes, spend arbitration, vendor/model currency review), a monthly human review of the findings ledger, and quarterly evidence packs mapped to NIST AI RMF, ISO 42001, and the EU AI Act.
$5–15K/month
Ongoing
Questions
Does Standing Watch replace our SAP, Workday, or Databricks governance tools?+
No. Standing Watch uses those tools' discovery, risk-rating, and observability data as inputs where you can export them, and adds what a single-platform tool isn't positioned to do across vendor boundaries: policy translation to NIST/ISO/EU AI Act, cross-vendor spend arbitration, and behavioral verification that doesn't stop at a closed platform's edge.
Is this actually running somewhere, or is it a framework on paper?+
It's running code Tioga operates on its own infrastructure today, and I'll show it in a demo: the router registry, a dated router-watch proposal report, and the security-watch findings ledger. It's honestly personal-scale — a two-machine, five-backend estate, not a Fortune 500 deployment — and I say so before you have to ask.
How is this different from a governance platform we could just buy?+
A platform gives you a pane — often a good one, and you may well buy one. It records what its own estate reports about itself. Discovering an agent inside a closed platform doesn't by itself establish that its actions were authorized, valid under that platform's own logic, and consistent with your controls, and a platform vendor has its own interest in spend or workload decisions between itself and its competitors. Standing Watch is a cross-platform, fixed-cadence verification and evidence layer that sits on top of any platform you already own. We review across platforms, and we disclose that our own agent-building work is built on Claude: client material is processed with Anthropic models by default, and a second provider only with your written approval in the statement of work. Tioga discloses its own vendor relationships, including that it may use credits from Anthropic's startup program.
Is a review independent of what Tioga builds?+
Reviews and builds are separate. Review findings are delivered before any build is scoped or priced, and you can take them to any builder, including one that competes with Tioga. A build is its own statement of work with its own acceptance tests. Tioga gives no opinion on a control it built beyond those tests, and does not call a control it built independently verified by Tioga.
How does the engagement actually progress?+
Assessment first — a fixed-fee diagnostic that seeds a real findings ledger you keep regardless of what you decide next. If it's a fit, Build implements the probe harness and review workflow in your environment, on top of the control plane you already use. The Retainer then runs it weekly, going forward, as your team's own control owners review and approve every change.