WorkingAgents + Alembic: Agent Governance Meets Causal Intelligence

By James Aspinwall, co-written by Alfred Pennyworth (my trusted AI) — March 7, 2026, 21:22

Alembic tells enterprises why things happened. WorkingAgents governs the agents that make things happen. When autonomous AI agents start executing marketing campaigns, adjusting budgets, and triggering omnichannel workflows, the causal intelligence that measures their impact needs a governance layer that controls their actions. That intersection is the partnership.

What Alembic Brings

Alembic is a Causal AI platform backed by $145M in Series B funding (led by Prysm Capital and Accenture, 15.7x valuation increase over Series A). NVIDIA is the founding enterprise customer and exclusive supercomputing partner. Fortune 500 clients include Delta Air Lines and Mars.

Core Technology:

Spiking neural networks (SNNs) — brain-inspired, patent-pending causal algorithms that fire in pulses like real neurons, processing only when needed
Spatio-temporal graph construction — every marketing event (campaign, sponsorship, media appearance) becomes a node, with edges representing directional causal influence
Counterfactual simulation — millions of “what if” scenarios isolating what actually caused an outcome vs. what merely correlated
Continuous learning — fresh forecasts in days, not quarters of historical data

What it delivers:

Deterministic attribution connecting ad spend to revenue — not correlation, causation
30-60-90 day predictive modeling across brand, performance, and omnichannel budgets
Anomaly detection and baseline observability
SOC 2 compliant, no PII/cookie collection, GDPR ready

Infrastructure: Private NVIDIA DGX AI Supercomputing cluster (NVL72 superPOD) at Equinix San Jose, processing billions of anonymized data points in real time.

Strategic partnerships: Accenture (integration into client transformation initiatives), NVIDIA (compute + algorithmic collaboration), Equinix (AI-ready infrastructure).

What WorkingAgents Brings

WorkingAgents is the access control layer and orchestration engine between AI agents and the systems they interact with. Three gateways — AI Gateway (250+ LLM routing), Agent Gateway (multi-step workflow orchestration), MCP Gateway (permission-scoped tool access) — with capability-based access control, guardrails at every checkpoint, and audit trails on every action.

Where the Synergies Are

1. Governed Agents That Act on Causal Insights

Alembic’s causal graph identifies that a specific campaign is driving revenue. Today, a human reads that insight and manually adjusts budgets. Tomorrow, an AI agent does it autonomously.

That agent needs governance:

Marketing Optimization Agent
  ✓ Alembic: read causal attribution data
  ✓ Alembic: read 30/60/90-day forecasts
  ✓ Ad platforms: adjust budget within approved range (±15%)
  ✓ Notifications: alert marketing director of changes
  × Ad platforms: budget changes exceeding ±15% (requires human approval)
  × Alembic: modify causal models or attribution logic
  × Finance: access to non-marketing budgets

WorkingAgents’ Agent Gateway orchestrates the workflow — read the causal insight, validate the budget adjustment against guardrails, execute within approved bounds, log every action. The marketing director gets a notification: “Agent shifted $12K from underperforming display to high-causal-impact social — approve continuation?”

Alembic provides the intelligence. WorkingAgents provides the governed execution.

2. Causal Attribution for Agent Actions

Enterprises deploying AI agents through WorkingAgents need to answer: did the agent’s actions actually cause the outcome?

An agent sends a personalized campaign sequence. Revenue increases. Was it the agent’s campaign, the seasonal trend, a competitor’s misstep, or the PR hit that ran the same week? Correlation says the campaign worked. Alembic’s causal graph says whether it actually did.

WorkingAgents’ audit trail records exactly what each agent did, when, through which tools, at what cost. Feed that action log into Alembic’s spatio-temporal graph as nodes. Alembic’s counterfactual engine determines whether those specific agent actions causally drove the revenue — or whether the outcome would have occurred anyway.

This closes the loop: agents act (governed by WorkingAgents) → outcomes are measured (attributed by Alembic) → future agent behavior is adjusted based on causal evidence, not correlation.

3. Cross-System Orchestration for Omnichannel Campaigns

Alembic ingests data from POS systems, ad platforms, social/web analytics, TV, radio, podcasts, trade shows, and third-party media. But ingesting data and acting on data are different problems.

WorkingAgents’ MCP Gateway connects agents to those same systems for execution — not just measurement. An agent that can read Alembic’s causal graph and execute changes across ad platforms, CRM, email, and social through governed MCP tools creates a closed-loop optimization cycle:

Alembic identifies causal drivers across channels
Agent proposes budget reallocation based on causal evidence
WorkingAgents validates the action against permission boundaries and guardrails
Agent executes across platforms (within approved scope)
Alembic measures the causal impact of the change
Cycle repeats with accumulated causal evidence

Each step is permissioned, logged, and auditable. The agent can’t overspend, can’t access unauthorized channels, and can’t modify attribution logic.

4. Compliance and Audit Trail for AI-Driven Spend

Marketing teams spending millions on AI-optimized campaigns will face scrutiny — from CFOs, boards, and regulators (EU AI Act high-risk provisions, August 2026). Two questions arise:

What did the agent decide, and why? WorkingAgents’ audit trail answers this — every action, every tool call, every guardrail evaluation.
Did the agent’s decisions actually produce results? Alembic’s causal attribution answers this — deterministic, not correlative.

Together, they provide a complete accountability chain: decision → action → governed execution → causal outcome measurement. Finance teams can trace every dollar from agent decision to attributed revenue impact.

5. Causal Intelligence Beyond Marketing

Alembic’s ambition extends beyond marketing attribution to causal inference across the enterprise. WorkingAgents is positioned for the same expansion — governing agents across any business domain.

As Alembic expands into supply chain, operations, and financial planning, WorkingAgents provides the governance layer for agents that act on those causal insights: supply chain agents adjusting orders based on causal demand signals, operations agents reallocating resources based on causal bottleneck analysis, financial planning agents adjusting forecasts based on causal revenue drivers — each governed with role-appropriate permissions and audit trails.

Partnership Model

Phase 1 (Weeks 1-4): Wrap Alembic’s platform as MCP tools (alembic.attribution, alembic.forecast, alembic.anomaly_detect). Define permission scopes — who can read insights vs. who can trigger automated actions. Prototype a governed budget optimization workflow.

Phase 2 (Weeks 5-10): Feed WorkingAgents’ agent action logs into Alembic’s causal graph. Measure whether governed agent actions causally drive outcomes. Build the closed-loop optimization cycle.

Phase 3 (Weeks 11-16): Joint go-to-market targeting Alembic’s Fortune 500 clients. Package “Causal AI + Governed Execution” for enterprises that want AI agents optimizing spend with deterministic attribution and full audit trails.

Why This Works

Alembic answers the hardest question in enterprise AI: did this actually cause that? WorkingAgents answers the second hardest: who authorized this, what happened, and can we prove it?

Causal intelligence without governed execution is insight without action. Governed execution without causal measurement is action without accountability. Together, they deliver AI agents that act on proven causal drivers, within defined boundaries, with a complete chain from decision to outcome.

James Aspinwall is the founder of WorkingAgents, an AI governance platform specializing in agent access control, security, and integration services for enterprises deploying AI at scale.