DcisionAI

The Intelligent Enterprise decision layer
DcisionAI
From knowing to deciding — the infrastructure layer that closes the gap
Every team. Every decision. The answer it deserves.
The Gap Between Knowing and Deciding
Every enterprise runs on data. That data feeds models. Those models power dashboards. And yet the actual decision — what to do, under real constraints, with real consequences — still falls outside the current system. No layer was built to own it.
That gap between knowing and deciding is where value is lost, where the best analysis in the room has no path to a credible, compliant solution. The problem is not shortage of data, and it is not shortage of judgment. It is a structural one: the decisions that matter most have outrun the tools available to make them.
Decisions requiring simultaneous satisfaction of dozens of constraints, competing objectives, and millions of possible actions have no adequate tool — until now.
80 Years of Science. Finally Accessible.
Operations research was formalized during World War II — convoy routing, logistics, resource allocation under constraint. Its methods underpin some of the most consequential decisions in modern business.
Asset Management
Portfolio optimization underpins capital allocation at the world's largest asset managers.
Airline Pricing
Yield management systems determine the price of every airline seat you have ever bought.
Supply Chain
Supply chain models kept grocery shelves stocked during a global pandemic.
"You have to somehow make high-quality, high-velocity decisions. It's easy for startups and very challenging for large organizations."
— Jeff Bezos, Amazon Shareholder Letter, 2016
Bezos called it nearly a decade ago. Most enterprises still haven't solved it. The problem is not awareness or the science — it is infrastructure. No layer was ever built to own the decision itself. DcisionAI is that layer.
How DcisionAI Works
A user describes a problem in plain English. DcisionAI routes it through a six-agent pipeline — Discovery, Research, Planning, Model, Solve, Explain — with mathematical audit gates that verify correctness before any output reaches the user. The result is not a recommendation. It is not a prediction. It is a certified optimal decision: the best possible answer given the rules of the problem, with every binding constraint named in business language and every shadow price expressed in dollar terms.
The solver returns not just an answer but a proof — the optimal solution, the constraints binding it, and the shadow price of each one: what it would be worth, in dollars, to relax it.
For the Technically Curious
What's inside every certified decision — the five signals the solver returns on every run.
Binding Constraint
A rule that is actively limiting the optimal solution. If you relaxed it, the answer would improve. The agent tells you exactly which constraints are binding — and by how much.
Business translation: This is the rule that is costing you money right now.
Shadow Price
The dollar value of relaxing a binding constraint by one unit. Not an estimate — a mathematically derived value,
Business translation: Relaxing this ESG concentration limit by 2% is worth $1.8M in this portfolio. Now it is a business decision, not a compliance checkbox.
Slack
The unused capacity in a non-binding constraint — how much room exists before a rule becomes limiting. Slack tells you where you have flexibility.
Business translation: You have 12% of headroom on this constraint. You are leaving optionality on the table.
Infeasibility
When no solution satisfies all constraints simultaneously. The problem has no answer as stated. This is not a failure — it is the most valuable diagnostic the solver can produce.
Business translation: Your rules are in conflict. The solver found it in seconds.
IIS — Irreducible Infeasible Set
The minimum set of constraints that together make the problem unsolvable. Remove any one of them and a solution exists. The IIS pinpoints exactly which rules are fighting each other.
Business translation: It is not that your problem is impossible. It is that these two specific rules cannot coexist. Now you know exactly what to negotiate.
When the Math Finds What Humans Miss
Infeasibility is not a failure — it is the most valuable diagnostic DcisionAI produces. Here is what happens when constraints collide.
1
Problem Submitted
The manager describes the problem in plain English: allocate $240M across 38 positions subject to ESG mandates, sector concentration limits, and a minimum yield requirement.
2
DcisionAI Finds Infeasibility
No solution satisfies all constraints simultaneously. The solver agent does not return an error — it returns an IIS: the ESG mandate and the minimum yield requirement are in direct mathematical conflict.
3
Shadow Price Surfaces the Tradeoff
Relaxing the minimum yield requirement by 30 basis points resolves the infeasibility. The shadow price quantifies it: that relaxation is worth $2.1M in portfolio value. The math is on the table.
4
Human Reviews and Overrides
The manager reviews the IIS and the shadow price. She decides to relax the yield floor — but only for two positions, not the full portfolio. She documents her rationale: client risk tolerance updated in last review.
5
Override Captured, Context Graph Updated
Who overrode: the manager. What changed: yield floor relaxed on positions 14 and 27. Why: client risk tolerance. Under what constraints: ESG mandate held firm. This is now institutional knowledge — encoded, auditable, and available on the next run.
The diagnostic took seconds. The override took minutes. The context — the IIS, the shadow price, the human rationale — is now part of the organization's decision graph. The next time a similar problem runs, it starts smarter.
The DcisionAI Learning Flywheel
A user describes a problem in plain English. The platform routes it through a six-agent pipeline with mathematical audit gates — and every run, every override, every outcome feeds back into the system, making the next decision faster and smarter. Three signal types feed the flywheel. Each loop is cheaper and more accurate than the last — and the asset it builds is organizationally proprietary.
All DcisionAI Runs Feed One Asset: The Context Graph


Agent Signals
What the agents found — optimal solution, binding constraints, shadow prices, conflict isolation. Recorded on every run.

Infeasibility Signals
Where the math found a conflict — infeasible states, constraint collisions, impossible tradeoffs which humans can override. Every infeasibility compounds into the graph.

Human Signals
Where humans disagreed with the math — who overrode, why, and under what constraints. The signal that builds institutional memory.
The flywheel's moat is specificity. The constraint logic, the failure patterns, the human judgment encoded over time. Every loop tightens. Every run is worth more than the last.
The Demand Is Structural
This is not a compliance story. Compliance is the floor. The structural demand drivers are larger, older, and more urgent than any regulation.
Decision Velocity Has Outrun Human Capacity
77% of business leaders report making more high-level decisions than a year ago — with 55% experiencing decision paralysis from sheer volume. (Pleo Decisions Report, 2025). 88% of enterprises have implemented or plan to pilot decision intelligence initiatives to close the gap between insight and action. (IDC / Aera Technology, 2025). The bottleneck is not data or talent. It is throughput.
The Cost of a Wrong Decision Has Compounded
McKinsey and the Institute of Directors estimate inefficient decision-making costs a typical Fortune 500 company $250 million per year in lost value. 77% of UK business leaders cite rising business complexity as the top driver of decision overload — with stress, paralysis, and missed opportunities as direct consequences. (Pleo, 2025). In today's margin environment, the gap between a good decision and an optimal one is measurable in dollars, not basis points.
The Expert Dependency Problem
Fortune 500 companies lose $31.5 billion annually due to institutional knowledge walking out the door. (SHRM). 51% of employees are actively seeking new opportunities in 2025 — each departure taking tacit decision context that rarely gets documented. (Work Institute, 2025). When the consultant leaves, the context leaves with them. The context graph solves this structurally — institutional decision logic becomes an asset, not a person.
Compliance & Auditability
The EU AI Act is fully enforceable from August 2026 — with fines up to €35 million or 7% of global annual turnover. High-risk AI decisions must be explainable, auditable, and documented. Third-party conformity assessments cost €15,000–€100,000 per system. (EU AI Risk, 2025). DcisionAI's certified outputs and context graph are built for this standard from day one. This is the floor, not the ceiling.
The enterprises that will define the next decade are not the ones with the most data or the most dashboards. They are the ones that close the gap — that turn every decision into a certified, optimal, auditable result. Compliance accelerates the urgency. The structural demand was already there.
Salesforce Integration
Why This Matters Inside Salesforce
Salesforce is the system of record for 150,000+ enterprises. For financial services firms — RIAs, PE/VC funds, private credit managers, wealth advisors — it holds the deal pipeline, the client portfolios, the account relationships, and the compliance records. But Salesforce has never had a decision layer that operates on that data under real constraints.
Today: The Advisor's Workflow
01
Export the data
Pull portfolio positions, client preferences, and constraints out of Salesforce into a spreadsheet.
02
Build the model
Manually construct allocation logic across 14 model portfolios, ESG rules, tax constraints, and concentration limits.
03
Make the judgment call
Hope no constraint was forgotten. Document the rationale manually. Pray the Reg BI trail holds up.
With DcisionAI as an Agentforce Agent
01
Pull data directly
Portfolio positions, constraints, and client preferences flow from Salesforce objects via MCP — no export, no CSV, no manual entry.
02
Run certified optimization
The solver returns the provably optimal allocation, names which constraint costs the most, and flags every binding rule.
03
Generate the audit trail
Reg BI documentation is produced in the same run. The entire decision trace is stored — without leaving the Salesforce ecosystem.
A spreadsheet can produce an allocation. It cannot prove optimality, surface which rule costs the most, or generate a defensible audit trail.
Architecture
How It Maps to Headless 360
DcisionAI's architecture was built MCP-native from day one. The six-agent pipeline already runs as MCP tools on FastMCP. Headless 360 does not require DcisionAI to change its architecture — it requires DcisionAI to publish it. Every layer of the Salesforce platform maps to a component DcisionAI already runs.
Data 360
Replaces manual data ingestion entirely. Discovery and Research agents pull deal pipeline data, portfolio positions, and constraint parameters directly from Salesforce's data layer via MCP. The constraint graph is populated from live CRM data — not user-typed inputs.
Experience Layer
The certified allocation, constraint economics table, and exclusion rationale render as native interactive components inside Slack, mobile, or any MCP-compatible surface — as decision tiles, right where the investment committee works. Not a PDF. Not a dashboard tab.
Agent Fabric
DcisionAI's sequential pipeline — with hard stage dependencies and mathematical audit gates between each stage — maps directly to Salesforce's deterministic orchestration model. The agents inherit Salesforce's trust layer, permission controls, and compliance guardrails.
AgentExchange
DcisionAI publishes as an MCP server on AgentExchange. Financial services firms discover it through AI-guided semantic search. One-click activation. No managed package. No installation overhead. Distribution at platform scale from day one.
Live Example
What a Customer Run Looks Like
A private credit fund — $380M deployed, $95M remaining — needs to allocate across 8 pipeline deals. Here is exactly what happens, from Salesforce data to certified decision, in five steps.
1
Step 1 — Data Ingestion
The Agentforce agent calls DcisionAI. Data 360 feeds deal attributes (yield, sector, risk score, climate tag, max deal size) and fund constraints (yield floor, climate mandate, sector caps, risk ceiling, single-deal cap) directly from Salesforce objects. No export. No CSV.
2
Step 2 — Pipeline Execution
Discovery identifies the problem type. Research enriches with market context. Planning formulates the mathematical model. The solver executes. The Explain Agent produces the decision intelligence report — five signal types, fully annotated.
3
Step 3 — Certified Result
Deploy $95M across five deals at 11.5% weighted yield. Three deals excluded with explicit rationale. Toll Road needs 50 bps improvement to enter. The deployment constraint binds at $0.328 per additional $1M. The certificate of optimality proves no feasible allocation produces a higher yield.
4
Step 4 — Interactive Review
Results render inside Slack as interactive components. The investment committee reviews the allocation table, constraint economics, and exclusion ledger. Want to raise the risk ceiling? Adjust and re-run. A new certified answer returns in seconds — not hours.
5
Step 5 — Institutional Memory
The decision trace stores in the constraint graph. Who decided. What was optimal. Which constraints bound. What was overridden and why. The next quarterly deployment starts smarter. The constraint graph remembers what the spreadsheet forgot.
Go-to-Market
Where We Start: Financial Services
High-constraint. Audit-required. Asymmetric consequences. We chose financial services not because it is the only domain — but because it is the hardest one. Win here, and every other domain is easier. Three beachheads, one mathematical substrate.
RIA / Wealth Management
The decision: Allocate capital across client portfolios subject to ESG mandates, concentration limits, tax constraints, and Reg BI documentation requirements.
Why now: Every advisor needs a defensible, documented rationale for every allocation. Reg BI requires it. None of the current tools in the Salesforce ecosystem produce one. DcisionAI generates the certified allocation and the audit trail in a single run.
Sample Run — “Multi-account household optimization”: A household holds $4.2M across taxable, IRA, and Roth accounts. The advisor must allocate across 14 model portfolios subject to client ESG preferences, asset location rules, single-position concentration under 5%, tax-loss carry-forwards, and Reg BI documentation. The platform returns the optimal allocation and names the binding constraint — asset location is binding; moving $180K of corporate bonds from taxable to IRA is worth $11,400/year in after-tax yield — and produces the Reg BI rationale automatically.
PE / VC / Private Credit
The decision: Allocate capital across portfolio companies, structure fund deployment, and optimize pipeline selection under return constraints, sector limits, and LP mandates.
Why now: Deployment decisions are infrequent but asymmetric — a wrong call compounds over a 10-year fund life. When the partner leaves, the decision context leaves with them. The constraint graph solves this structurally, preserving institutional memory that currently exists nowhere.
Sample Run — “Capital deployment in a private credit fund”: A $380M EM private credit fund must deploy $95M across 8 pipeline deals, subject to a 10.5% yield floor, a 40% climate mandate, sector concentration caps, a 4.0 risk ceiling, and a 30% single-deal cap. The platform returns: deploy across 5 deals at 11.5% weighted yield. Three deals excluded with explicit reduced costs. The deployment constraint binds at a shadow price of $0.328 per $1M. The certificate of optimality proves no feasible $95M allocation produces a higher yield-weighted result.
Fund Administration
The decision: Calculate waterfall distributions, structure LP allocations subject to preferred return and catch-up provisions, and produce audit-ready compliance documentation.
Why now: When a fund's LPA language is ambiguous — deal-by-deal or European carry mechanic? — a spreadsheet produces a number under whichever interpretation was hardcoded. DcisionAI's pre-solve gate flags the ambiguity, computes distributions under both interpretations, and captures which interpretation was chosen, by whom, and why.
Sample Run — “Waterfall calculation under structural ambiguity”: A fund's LPA language is ambiguous on whether the carry mechanic is deal-by-deal or European. A spreadsheet cannot detect the ambiguity — it produces a number under whichever interpretation was hard-coded. The platform's pre-solve gate flags the ambiguity, surfaces both interpretations, and computes LP-level distributions under each — under deal-by-deal, GP catch-up triggers in year 3; under European, year 6. The timing difference is $4.1M across this LP cohort.
Expansion
The Wedge: Same Infrastructure, Every Domain
The three financial services beachheads share one mathematical substrate — constrained optimization with auditability requirements. That substrate is domain-agnostic. Once the pipeline is proven and the constraint graph is seeded, expansion requires no new infrastructure — only new problem types.
Financial Services
Portfolio allocation, fund deployment, waterfall distribution, wealth advisory — the founding vertical. Maximum constraint complexity. Maximum audit requirement.
Healthcare
Resource scheduling, staff allocation, capacity optimization — same constrained optimization engine, applied to patient flow and clinical resource constraints.
Logistics
Routing, load optimization, last-mile delivery — the discipline that DcisionAI's solvers already power at the world's largest logistics networks.
Supply Chain
Inventory allocation, supplier selection, procurement optimization — multi-echelon constraint problems that currently live in spreadsheets and off-platform tools.
For Salesforce: a single AgentExchange listing that serves every vertical where customers face complex, constraint-bound decisions — which is every vertical Salesforce serves.
The Opportunity
A Category That Does Not Yet Exist on the Platform
For Salesforce Customers
A decision layer that operates on the data they already have, under the constraints they already face, producing certified optimal answers with full audit trails — without leaving the Salesforce ecosystem. The constraint graph compounds as institutional knowledge alongside their CRM data.
For the Salesforce Ecosystem
A category-defining capability that no current AgentExchange listing provides. Not another dashboard. Not another AI assistant. A mathematical optimization engine that proves the best possible answer — and stores the decision trace as compounding proprietary knowledge.
For the Agentforce Story
A concrete, high-value demonstration of what agents can do when they have access to real business data, real constraints, and real solvers — work previously locked behind six-figure consulting engagements and months of implementation. Now available in a single Agentforce activation.
DcisionAI does not compete with anything in the Salesforce ecosystem today. It creates a category that does not yet exist on the platform: the certified decision layer.
Every team. Every decision. The answer it deserves.
DcisionAI is the infrastructure layer that makes the next decade of enterprise possible. The compliance environment is accelerating the urgency. But the opportunity is larger than regulation. It is the permanent elevation of how organizations decide.