Case study: an AI energy audit tool

Return on investment

100Consultant days/year savedEquivalent to roughly £120k of billable capacity.

£30kUpsold from the same dataAdditional deliverables, such as Carbon Reduction Plans, sold from data already collected.

Growth without headcount

Fourteen additional SECR clients and eleven new ESOS engagements taken on in 2026, without adding headcount.

Background

A growing consultancy, limited by its own data work

Our client is a twelve-person energy and carbon consultancy working with mid-sized manufacturers, logistics operators and multi-site retailers across the UK. Two of its consultants are registered ESOS Lead Assessors; the rest are energy analysts and carbon accountants. The firm delivers around forty SECR disclosures a year, supports thirty clients through ESOS, and runs a growing carbon strategy practice covering SBTi target-setting, net zero roadmaps and Carbon Reduction Plans. The expertise was never the constraint. Getting the data into a state where that expertise could be applied was.

Offshore wind turbine rising above low cloud

The challenge

Two Lead Assessors spending their time on data, not judgement

With its ESOS pipeline building and its SECR book expanding, the firm was turning work away because its two Lead Assessors were spending most of their time on data rather than judgement. Hiring another Lead Assessor would have taken a year and still left the underlying process unchanged.

01

Data work crowding out judgement

Chasing utility bills, reconciling meter exports against invoices, and rebuilding the same consumption dataset in a fresh spreadsheet for each scheme.

02

Every client sends something different

Supplier PDFs, photographs of bills, half-hourly exports, fuel-card statements, mileage claims, F-gas registers and waste notes. No two clients send the same formats, and someone had to read every one of them.

03

The same data, collected twice

A client's SECR data and its ESOS data were collected twice, by different people, in different formats, and the two rarely agreed at the first pass.

04

Hiring wouldn't fix the process

Hiring another Lead Assessor would have taken a year and still left the underlying process unchanged.

Mainspring solution

AI does the reading. The assessor does the deciding.

Whatever arrives, read and mappedSupplier PDFs, bill photos, half-hourly exports, fuel-card statements and mileage claims are read as they arrive and mapped to the right site, meter or Scope 3 activity. Duplicates across formats are caught, and the coverage grid shows month by month what is real, estimated, missing or contradictory.
Anomalies explained, not just flaggedWhere a reading looks wrong the platform says why: an estimated read where a meter read was expected, an invoice overlapping the previous period, a half-hourly export disagreeing with the bill. The analyst resolves rather than investigates.
Drafted in the firm's voice, held until signedNarratives, savings opportunities and measures are drafted from the client's own consumption profile, sector and history, under the firm's branding. Every figure and paragraph carries a draft label until a named Lead Assessor marks it reviewed.

See it in actionPick a screen, use the arrows, or tap the cards behind

01Scope tested, not assumed02Everything the client sends, read on arrival03One dataset behind every report04Twelve stages, each one decided

01 / 04

02 / 04

03 / 04

04 / 04

01

Scope tested, not assumed

Every client profile runs through deterministic tests for ESOS, SECR, PPN 006, CCA, TCFD, SBTi and ISO 50001. Each verdict carries its reason and the rule behind it, and re-evaluates the moment staff, turnover or balance-sheet figures change. Where the rules do not reach, the regulations themselves can be questioned directly and answered with the clause cited.

ESOS · SECR · CRPRule shownAsk the regulation

app.mainspring.ai/clients

Six clients screened against seven regimes at once, each verdict carrying its reason.

02

Everything the client sends, read on arrival

Half-hourly exports, supplier bills, bill photos, fuel-card statements, mileage claims, F-gas registers and waste notes go in as they arrive. Each file is read, parsed and mapped to a site, meter or Scope 3 activity, and the coverage grid shows month by month where data is real, estimated, missing or contradictory.

Any formatRead and mappedGaps explained

app.mainspring.ai/data-sources

Everything the client sends, in one place

A year of bills, meter data and mileage reconciled into one grid — before a single number is calculated.

03

One dataset behind every report

Energy in kWh and emissions by scope are held once per client, and ESOS, SECR and the Carbon Reduction Plan are each a view of it. Scope 3 categories are labelled with the regime that requires them, so the same activity data satisfies several filings without being rebuilt.

Scope 1, 2, 3Regime-taggedNo rekeying

app.mainspring.ai/emissions

One set of numbers, three filings, with each Scope 3 category tagged to the regime that requires it.

04

Twelve stages, each one decided

ESOS runs as twelve stages, from qualification through the 95% coverage test to sign-off. The platform drafts measures and narrative from the client's own records, but each stage stays open until the Lead Assessor records a decision and the basis for it.

12 stagesDrafted, not decidedDecision logged

app.mainspring.ai/esos/report

Seven of twelve stages decided — the report cannot be filed until the rest are.

The outcome

Capacity released, and judgement back where it belongs

The firm took on more work, sold more from the same data, and kept every decision with a named assessor:

A hundred consultant days a year, returned. Roughly £120k of billable capacity no longer spent reading bills and rebuilding spreadsheets.

Growth without hiring. Fourteen more SECR clients and eleven new ESOS engagements in 2026, on the same headcount, with no new Lead Assessor needed.

More sold from the same data. £30k of additional deliverables, including Carbon Reduction Plans, from data already collected rather than gathered again.

The assessor's name on every decision. Drafts are generated, never issued. Each figure and paragraph carries a draft label until a named Lead Assessor marks it reviewed.

Work with us