A $1.2 billion, five-year capital plan heads into a commission that disallowed 8% of the last one. Watch GridCORTEX score all 240 candidate projects by risk-reduction-per-dollar, expose the plan sitting 23% below the efficient frontier, and (with a human approving every move) reallocate the same budget onto the frontier, then generate the testimony-ready evidence pack that survives cross-examination. Same dollars. More risk retired. Zero disallowed.
| Without | With GridCORTEX | Δ |
|---|
| Without | With GridCORTEX | Δ |
|---|
The scenario follows a fictional utility taking a $1.2 billion, five-year infrastructure spending plan into its state utility commission, the regulator that decides whether customers' rates can pay for that spending. Last time, the commission refused to let 8% of the plan be recovered from customers (a "disallowance", which shareholders absorb), and it has warned it will look hard at projects justified only by equipment age. The plan is 240 candidate projects gathered from nine departments' wish lists, and the simulation runs 16 weeks, from assembling the plan to the commission's final ruling. In Week 1, GridCORTEX reads the working computer models of the utility's real equipment (fed by oil tests that reveal a transformer's internal health, loading data, and inspections), ten years of outage records, wildfire and flood exposure by location, and demand growth. It then scores every one of the 240 projects the same way: how much risk does each dollar spent actually remove? The answer lands in Week 1.6: the current plan removes 61 risk points, but the same $1.2B, allocated to the best projects, would remove 75. The plan is 23% below the best available line, meaning nearly a quarter of its risk-reduction potential is being wasted. The cause has a name: 38 funded projects replace equipment purely because it is old, even though its condition data says it is healthy, while the territory's genuinely riskiest equipment sits unfunded.
Two decision points ask for human approval. The first, around Week 3, is the reallocation: 38 age-only projects out, 51 higher-risk projects in (burying lines in wildfire zones, rebuilding the worst-performing local circuits, and 12 new tie switches, which let a neighborhood be fed from a second direction when its usual line fails). Same budget, and every swap carries a written justification. Approved, the portfolio moves from 213 to 226 projects. The projected payoff: the average customer's outage time drops 9 minutes per year, and the modeled chance of utility equipment starting a wildfire drops 31% in the highest-risk zones. The second decision, around Week 5, is the evidence pack: 214 per-project exhibits totaling 1,900 pages, a methods document written for the hearing room, every number traceable back to the sensor reading or record behind it, and pre-drafted answers to the 40 questions this commission has historically asked, mined from its past cases.
Then the case plays out. The plan is filed in Week 6. In Week 8, 47 formal written questions arrive from the commission staff and outside challengers; they are answered from the same model in 4 days, where the same job historically took six weeks across nine departments. At the Week 11 hearing, a staff engineer cross-examines the risk methods for three hours, and every challenged number traces cleanly to an equipment reading, an outage record, or a work order. The ending, in Week 14: the plan is approved with zero dollars disallowed, the ruling citing "a transparent, quantified risk methodology." Skip the approvals and the same case ends in Week 15 with $96 million disallowed, because "the equipment was old" did not survive cross-examination.
The plan is nine departments' lists stapled together, never scored on one common risk scale, with equipment age standing in for actual risk and filing numbers living in spreadsheets that disagree with each other. Consumer advocates and other outside challengers attack the 38 healthy-equipment replacements as wasteful spending, "replacement by birthday, not by risk," and the utility's rebuttal can only cite engineering judgment, because the numbers were never computed. The commission staff recommends cutting 8%, and the ruling takes $96M out of the plan. Shareholders absorb the loss, the reliability work that money funded slips a full multi-year cycle, preparing the filing costs 3,800 staff-hours, answering the formal questions takes 6 weeks across 9 departments, and the next rate request starts from a credibility hole.
The software reads the equipment condition data, outage history, and hazard maps, scores all 240 projects on one risk-per-dollar scale (use case UC 14.2), and computes the best possible allocation of the same budget. Equipment lifetime economics (UC 4.4) show which age-based replacements do not earn their slot, and the question-response factory (UC 14.4) answers all 47 formal questions in 4 days from the same model that built the plan, so the filing and the evidence can never disagree. Humans decide everything: the planning committee approves every project swap, and the regulatory affairs team owns the filing. The result: the same $1.2B removes 23% more risk, filing preparation drops to 940 staff-hours, cross-examination finds no gaps, and the ruling approves the plan with $0 disallowed.
| KPI | Without GridCORTEX | With GridCORTEX | Delta |
|---|---|---|---|
| Risk retired ($1.2B)how much risk the $1.2 billion actually removes, scored on one comparable scale | 61 index pts | 75 index pts | +23% |
| Plan positionwhere the plan sits against the best possible risk-removed-per-dollar allocation | 23% below frontier | ON the frontier | no risk-reduction wasted |
| Projected SAIDI trajectorySAIDI is the industry measure of outage minutes the average customer suffers per year | flat | −9 min / yr | reliability up |
| Wildfire ignition risk (top zones)the modeled chance that utility equipment starts a fire in the highest-risk areas | baseline | −31% | resilience up |
| Healthy-asset replacements fundedprojects that replace equipment because it is old, even though its condition data says it is healthy | 38 | 0 | spending follows risk only |
| Worst-risk unfunded projectsthe highest-risk fixes the original plan left out entirely | stay unfunded | 51 funded | gap closed |
| New tie switches into the networkswitches that let a neighborhood be fed from a second direction when its usual line fails | 2 | 12 | more ways to restore power |
| Capital disallowedspending the commission refuses to let the utility recover from customers; shareholders absorb it | $96M | $0 | −$96M |
| Filing preparationstaff time to assemble the regulatory filing and its evidence | 3,800 staff-hours | 940 staff-hours | −75% |
| Data request turnaroundhow fast the utility answers the formal written questions regulators and challengers ask | 6 weeks · 9 departments | 4 days · one model | −93% |
| Hearing exposurehow the plan holds up under live cross-examination in the hearing room | methodology challenged | every number traced | no gaps found |
| Settlement positionwhether the utility must negotiate its plan down to avoid a worse ruling | weak; staff recommends cut | none needed | full approval |
| Next-cycle starting pointthe reputation the utility carries into its next rate request | credibility repair | precedent set | trust carries forward |
| Reliability work slippedapproved reliability projects delayed because their funding was cut | one full cycle | none | customers win |
The safety effect is indirect and downstream, and it should not be counted in the business case. Better justified investments get approved, and some of those investments are the hardening and automation projects that remove field exposure. That is a real chain but it is long, and the honest position is that the drafting work itself changes nobody's exposure.
Counted in units you already track:
Drafting and reconciliation hours come back to capital planning, and the regulatory writers stop spending their time translating and start spending it on framing and strategy.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Capital planning labor | drafting and reconciliation hours avoided x your loaded planner rate |
| Regulatory writing and review | translation and review hours avoided x your loaded regulatory writer and attorney rates |
| Consultant narrative preparation | consultant hours or fixed fees you stop paying for narrative work x your contracted rate |
| Rework from figure mismatches | correction cycles avoided x hours per cycle x the loaded rates of the people in the chain, plus the cost of any supplemental filing a mismatch would have forced |
| Capital timing | your own carrying cost of a project that slips a rate cycle because its justification was not ready. You supply the carrying rate and the probability. We will not estimate either. |
You pay for the scoped engagement that builds and runs this, for the integration that reads your project workbooks, asset records, and prior filings, and for the time your planners and regulatory writers spend verifying the first set of generated narratives line by line. Expect that verification to be slow at first and to get faster once the team trusts the traceability. Budget the first filing season as a review cost, not a savings.
Payback is normally driven by planner and regulatory writer hours plus any consultant narrative fees you stop paying. Capital timing is the biggest number in the model and the least defensible one, so present it separately and let your finance team decide whether to count it at all.
The mechanism is fatigue and diverted attention, not physical exposure, and we will not dress it up as anything else. Peak discovery runs the case team and its experts on sustained overtime for weeks, and several of those experts are operating and planning engineers whose day job is approving and scoping field work. Hours returned to them are hours available for the work that does carry exposure.
Counted in units you already track:
Research and drafting hours come back to the subject matter experts who are pulled off engineering and accounting work, and attorney time shifts from hunting for inconsistencies to judging the ones that are flagged.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Subject matter expert labor | research hours avoided x the loaded rate of the group the expert comes from, which is usually engineering or accounting, not legal |
| Attorney review | review and consistency checking hours avoided x your loaded in house rate and your outside counsel billed rate, split the way your case is actually staffed |
| Contract discovery support | temporary analyst weeks avoided x your agency rate |
| Rework and supplemental responses | corrections avoided x hours per correction x the loaded rates of everyone in the correction chain |
| Case exposure | your own estimate of what one inconsistent answer costs when it becomes cross examination material. Your legal team sizes this, not us, and many utilities choose to leave it out of the base case entirely. |
You pay for the scoped engagement that builds and runs this, for ingesting your testimony, workpapers, and the answer record from prior cases, and for the integration into your document management system. The largest cost you carry is your own experts' and attorneys' time to verify drafted responses closely during the first batches, until the team has calibrated how much checking each type of request actually needs. Do not plan to skip that stage.
Payback is carried almost entirely by subject matter expert hours and attorney review hours in a single active case, which is why utilities usually build this case on their last rate case and not on an annual average. Case exposure and rework are real but keep them as upside.
This one is indirect and should be described honestly: the optimizer never touches a work site. What it changes is how much of the fleet is deliberately allowed to run to failure, and every run to failure decision that turns out wrong becomes an emergency job. Fewer of those means fewer crews working unplanned, at night, on equipment nobody scoped in advance.
Counted in units you already track:
Plan assembly and scenario building hours come back to asset management and capital planning, and the director spends the season defending a plan instead of building one.
The numbers we need from you to run that formula:
| Cost driver | How it is calculated, from a rate you supply |
|---|---|
| Deferred capital | assets moved to a later year or to refurbishment x your installed cost per unit x your cost of capital x years deferred |
| Refurbishment versus replacement | units refurbished instead of replaced x the difference between your refurbishment cost and your installed replacement cost, net of the remaining life you are willing to credit a refurbished unit |
| Planning labor | plan assembly and scenario hours avoided x your loaded rate for asset management and capital planning staff |
| Risk weighted outage exposure | your cost per unplanned outage by asset class x the change in expected failures between the current plan and the ranked plan, computed with your own failure rates |
| Lead time exposure | units pulled earlier to beat a lead time x your cost of carrying that asset early, weighed against your own cost of being caught short |
You pay for the scoped engagement that builds and runs this, for integration into your capital planning and asset systems, and for the internal work of agreeing on inputs, particularly your cost per unplanned outage by asset class. That last item is a negotiation between asset management and finance and it usually takes longer than the integration does.
Payback is dominated by deferred and avoided capital rather than labor, because the dollars per decision are large. The credibility of the whole case rests on whether finance accepts your outage cost inputs, so settle those before you build anything.
What is this, exactly? It is AI software: intelligent agents and models built and delivered by SoftServe, running on NVIDIA accelerated computing. It is not a hardware appliance and it does not replace the systems you run today. It deploys in your own cloud or on your premises, connects read-only to your existing systems, and recommends; your people approve every action, starting in shadow mode until it earns trust.
A document generator for capital planning and regulatory teams. It takes project spreadsheets and produces regulatory-grade investment justification narratives, each traceable to your data and consistent with prior filed positions. The demo above uses synthetic data; everything below describes what the real deployment needs from your organization.
| Your system | Typical products | How we connect |
|---|---|---|
| Asset / work management (EAM/CMMS) | IBM Maximo, SAP PM, Oracle WAM | scheduled file export (CSV or CIM XML) |
| Planning and study tools | CYME, Synergi, PSS/E | scheduled file export (CSV or CIM XML) |
| Project scheduling | Oracle Primavera P6, Microsoft Project | scheduled file export (CSV or CIM XML) |
| Document and knowledge stores | SharePoint, OpenText | document upload |
| Financial and enterprise resource planning (ERP) | SAP S/4HANA, Oracle EBS | scheduled file export (CSV or CIM XML) |
| Customer Information System (CIS) / billing | Oracle CC&B, SAP IS-U | scheduled file export (CSV or CIM XML) |
| Regulatory docket sources | state commission e-filing portals | read-only API |
Runs in your own cloud account on GPU instances with read-only connections through your existing data zone. No system of record is changed; the output is a document your planners edit and own. The pilot generates justifications for three real projects alongside the manual versions.
The Approve button you just clicked in the demo above is the real workflow. This is what it looks like on the screen of a capital planning analyst in the GridCORTEX console:
Approve saves the narrative to the document store as a draft and routes it to regulatory affairs for expert edit through your existing review workflow. GridCORTEX files nothing; your team owns the final document.
The planner points at the project spreadsheet and picks the goal alignment (resilience, equity, or emissions); supporting data pulls automatically from EAM and planning tools.
Project data is read from EAM and planning tools at generation time and prior filings refresh daily; the draft shows the as-of date of every source table.
Lives in the GridCORTEX console with drafts filed in the document store; email push when a project narrative is ready. The console runs in a browser beside your existing screens on day one; embedding into your own systems is a roadmap step once the read-only phase has earned trust. Approve, Modify, and Decline are all captured in an audit trail your compliance team can pull, and GridCORTEX never blocks or overrides anything in the systems you run today.
The fair question from any utility planning shop: "We have an EAM, planning spreadsheets, consultants, and decades of rate-case experience, what's new here?" Here's the honest answer.
When someone asks "what did it actually calculate?", this is the list. In the simulation these factors drive the storyline; in a pilot they are computed from your EAM, asset twins, OMS history, GIS, and financial systems.
Presenter's one-liner: "It put all 240 projects on one risk scale, showed the plan sitting 23% below the efficient frontier, moved it there with the same dollars, and then wrote the evidence pack from the same model, so when the intervenors pushed, every number traced. Same budget, more risk retired, nothing disallowed. That's what you just watched."