A neuro-symbolic AI engine that surfaces evidence at six points of decision — improving services revenue and gross profit through better decisions, not new tools.
Every services engagement moves through lead, proposal, delivery. At each phase, decisions get made about revenue (what we expect to make) and cost (what we expect to consume). The cells are surfaces of intervention. The substrate is what makes them one system.
Today: pipeline reflects what sellers have qualified, not what the market has signaled. The engine watches signals continuously across both the existing customer base and the addressable market — surfacing where to focus, before the deal is in CRM.
Pre-sales architects, deal desk, executive sponsorship are scarce capacity. Today they get spread across all qualified pursuits. The engine surfaces opportunity cost per pursuit so the sales team allocates against expected portfolio-net value.
Today: static margin floor regardless of capacity state. The firm over-prices when capacity is available (loses winnable deals) and under-prices when capacity is scarce (commits scarce people to thin-margin work). The engine makes pricing capacity-state-aware.
Today: estimation by template + memory. Architects bid on a role-mix that's 5-10 points off what comparable engagements actually delivered at — baking in margin compression before the deal is signed. The engine puts the corpus inside the estimation tool.
Bookings aren't revenue. Engagements get paused, scope reduces, customers churn mid-delivery, recognition slips. Today these surface in finance reports after the fact. The engine builds a daily recognition trajectory from active engagement reality — control loop and forecast in one signal.
Today: project margin gets discovered at quarter-end variance review. By then the trajectory is locked. The engine produces a daily predicted-delivered-margin per engagement, with drift drivers attributed and intervention candidates surfaced — while intervention still works.
The cells aren't independent interventions. Delivery outcomes feed the corpus that calibrates next quarter's lead-scoring and proposal estimation. Lead-stage signals feed proposal-stage pricing context. Proposal-stage decisions set the baseline for delivery-stage drift detection. The engine learns continuously because the substrate is shared.
The engine's primary KPIs aren't financial — they're decision-quality measurements. They aggregate into intermediate metrics finance and sales already track, which aggregate into services revenue and services gross profit. Every dollar attributed to the engine is traceable down to specific rule firings on specific decisions.
Project services runs linear engagement cycles. Managed services runs a recurring renewal cycle. Staff augmentation runs a fast iterative loop. Same engine, three timeshapes — surfacing evidence in the operator's existing tool at the moment of decision.
The highest-dispersion service line. Four dominant cells: estimation accuracy and capacity-aware pricing at the front, in-flight steering and demand-shape signals at the back. Multiple intervention points, multiple compounding margin protections.
The largest absolute pocket. Cell 01 and Cell 05 dominate: continuous demand-shape signal + scope-true-up audit at renewal. Recurring revenue compounds — once captured, rebased MRR persists for the next term.
Smallest pocket as a percentage. Two dominant cells: bench-match scanning + capacity-aware bill-rate discipline. Per-placement margins are structurally thinner; the engine produces value through frequency, not magnitude.
One workspace, four panels, multiple cells firing concurrently. The architect doesn't switch contexts to get engine input — the PSA they already work in surfaces evidence inline. Bid pricing, engagement queue, notifications, decision log: every panel anchored to a specific cell, every recommendation logged with counterfactual.
Six primary KPIs, three intermediates, two outcomes. Every number on the screen traces to specific decisions logged in the operator UI. Confidence intervals are real (from holdout cohorts and counterfactuals), variance commentary writes itself, and override patterns surface where the engine needs recalibration.
Project services, managed services, and staff augmentation have different economic shapes — different cells dominate in each motion. The matrix below shows which cells are dominant, important, or minor for each service line. The deployment sequence follows from where the heavy levers concentrate.
The six cells don't move every service line equally. Where the heavy levers concentrate determines the addressable pocket. What the engine actually captures is bounded by adoption velocity, data discipline, and customer-acceptance dynamics — so we frame realistic capture (50%) and mature steady state (65%) separately.
The engine doesn't change the P&L's structure. It changes the lines — at realistic capture, on the firm's $600M services book. Each impacted line traces back to specific cells, which trace back to specific operator decisions.
No single next step. Four pathways — from internal pressure-test to engineering deep-dive to deployment. Pick the one that matches what you need to learn next.
Three phases · foundation, deployment, expansion. The GP First Instance covers the first 10 weeks — everything after ramps based on what the first instance reveals about data readiness and adoption.
The engine is inline, in the tools they already use. No new system to log into, no separate dashboard to remember to check. Adoption is a function of relevance and trust, not training.
Different in three specific ways: where it sits in the workflow, what it operates on, and what kind of attribution it produces. NSAI is not a replacement for any of these; it's a layer that connects them.
All operational data stays inside your firm's data perimeter. The engine deploys as a layer on top of your existing systems, not as an external service that exfiltrates data. Customer data, operator decisions, financial actuals: all in your environment.
Three roles matter: an executive sponsor, an operations lead, and a technical lead. The engine is operationally light to run once deployed; the work concentrates at deployment and at quarterly recalibration.