The agentic layer changes the
economics of fund administration.
Capacity that no longer requires matching headcount. Agentic effort and cost directed by risk. Quality evidenced as the work happens. Countersign is that layer — purpose-built for closed-end fund administration, deployed into your own cloud, under your control.
Agents do the heavy lifting · Your people deliver the judgement · Everything is evidenced
Countersign V1 is in build.
Why this is difficult — and what makes it work.
Designed to lower cost-to-serve with service quality maintained — efficiency that compounds each cycle.
Designed to absorb audit season and quarter-end, make smaller mandates economic to serve, and let growth run ahead of hiring.
Your agents and your staff, together, above the ledgers, document stores, portals and workflow you already use — one permission system for service delivery.
In-tenancy with no vendor-side dependency to operate — every deployed instance keeps running on your infrastructure. Continuity terms stand in every contract.
Model cost is incurred only where a model is called, on your own endpoint — a cloud enterprise agreement, a vendor service, or an open-weight model on your own infrastructure. Every call is recorded, so usage reads by client, process and model.
Every model call is recorded with its model, tokens and the item it served, so your provider's invoice can be checked against your own record — by client, process and model — rather than taken on trust.
Every item carries its own record of effort and model use, so cost-to-serve by client and by process is read from the operation as it runs, not reconstructed from timesheets and sampling.
Configuration drafted in pre-production with your own AI and confirmed by your team — not a multi-year programme.
Designed as a platform for consolidating and rationalising the systems in your estate over time.
Agentic capacity absorbs the volume while SLA timeframes and quality requirements hold — a platform for ongoing efficiency that learns through each cycle it runs.
Release guards that block, not warn: recipient must match the client, every document must belong to that client and fund, figures must reconcile.
Nothing replaced, self-served install — your test library and preferences grow in your tenancy. Countersign does not train models on your data; model terms follow your own agreement.
Installed in your own environment; documents stay in your stores; no second data estate. The InfoSec pack is provided before anything connects.
Outbound traffic limited to approved destinations; agents hold no send route; anything leaving the firm is released by a person, per item; inbound content treated as data, never instructions.
Pinned versions, per-workflow assignment. With GoldenEval and your risk profile, you approve the model(s) your work runs on — every change evidenced.
Models, costs and capabilities keep changing. Multi-model by design: swap, mix and re-assign by workflow and by risk as innovation lands — each change evidenced through GoldenEval.
You publish the specification your clients' AI conforms to — structured, credentialed, sandboxed. You lead that conversation.
Audit, SOC and resilience evidence assembles from records linked as the work happened, with sign-off — not reconstructed afterwards.
Thresholds and limits from your framework govern release paths; breaches route to staff and are recorded as exceptions.
Measures whether staff review is working — evidenced in the record.
Six steps. Your people keep the two that matter.
Model-agnostic, multi-model and model-flexible by design — different models, at different levels of effort, can be pointed at each step. Your risk rating can assign a stronger model to Check & Release.
Internal compliance, audit, SOC and resilience evidence assembled from records linked as the work happens — and, on the roadmap, audit-sample response as a triage workflow.
A flexible agentic core, combined with risk management.
Every platform says a person reviews the output; Countercheck measures whether that review is working — calibrated at onboarding, measured in operation, evidenced in the record.
Runs above the ledgers, mailboxes and stores you already keep; adopted team by team, at your pace.
Countersign sits across your mailboxes, stores, ledgers and portals, so whatever connects next is governed by the same permission system and recorded in the same record.
As your clients' AI arrives, you publish the specification it conforms to — credentialed, sandboxed, evidenced. The administrator leads the standard.
A model change sits your exam — your approved test set, marked against pass marks you set — and reaches live work when you activate it. GoldenEval-Live re-runs the exam on schedule. You hold the gate; the record holds the evidence.
Release is governed by your risk rating.
Every release is a staff decision — made per item, or, for low-risk internal work, governed in advance by a policy your firm sets and detailed control procedures. Anything leaving the firm is released by a person, per item. Nothing releases on an agent's own authority.
Countercheck measures whether staff review is working; GoldenEval-Live measures model performance.
Every item carries a risk rating from your firm’s framework — who checks, how it releases, which model runs and what assurance it inherits all follow from it. Boundary-crossing work stays a per-item staff release. Illustrative rows; every rating starts HIGH.
One framework drives it all: who checks · how it releases · which model runs · what assurance inherits. Implement by client, by team, by service, by country.
For Fund Administrators.
Countersign is built to deliver the value of agentic service efficiency — whilst maintaining service quality within your risk and control environment.
We are reaching out and working with a small number of fund administrators.