Roadmap

TSK has two distinct surfaces in production today, and we hold ourselves to honest coverage statements for each. The first is automated export coverage — how completely we can fill a machine-readable framework output from extracted evidence. The second is questionnaire-help coverage — which buyer questionnaires our supplier autofill surface understands. These are measured differently and should not be conflated.

The export registry pins coverage per framework adapter. CDP XML emits 3 of 13 modules (~23%). GRI CSV emits 6 of 31 numeric disclosures in the GRI 300-series environmental subset (~19% of that subset, not of total GRI which spans ~100+ disclosures across all pillars). PACT JSON is partial at the organisation level, not the product-LCA level. ISSB S2 and ESRS-VSME are advisory gap-analysis only — they emit a gap report, not a submission-grade artifact. CSRD iXBRL, TCFD narrative, and SASB industry templates are not implemented in any form.

The supplier autofill surface ships with three loaded templates: CDP Climate, EcoVadis, and GRI Universal. For each, TSK pre-fills quantitative slots from extracted evidence and prompts the supplier for qualitative slots, with source-attribution badges on every pre-filled answer. Broader buyer-ask templates are added on demand post-launch when a supplier needs one — that is not a coverage gap, it is an intentional on-ramp.

Underneath both surfaces, the pipeline does the same thing it has done since the beta: regex + LLM extraction, confidence-gated review, DEFRA 2025 emission factors, and a tamper-evident Supplier Pack ZIP at delivery. The /methodology page covers that engine in depth.

Honest limits

Scope 3 emissions are not implemented. Market-based Scope 2 is deferred — only location-based Scope 2 is supported today. Questionnaire autofill is 0% until a supplier uploads evidence; the metric is bounded by what the uploaded documents actually contain. The dual-LLM verification signal is logged internally but not yet surfaced in supplier-facing confidence badges — wiring that surface is a Next quarter item.

The next quarter is shaped by what we hear from suppliers and sustainability professionals during the launch window. The items below are planned but not committed — each is gated on signal we collect over the first weeks of public use. No dates, no guarantees. What we will not do is ship coverage numbers that we cannot stand behind during an audit.

The dual-LLM adjudication module already runs in production for extractions and logs disagreements internally for operations review. The next step is wiring the result into the confidence badges suppliers see on each extracted value — so a low-disagreement value shows a stronger badge than a high-disagreement value. The module exists; the supplier-facing surface does not.

More CDP modules and more GRI disclosures, prioritised by which values the launch cohort actually needs to disclose. We will not claim a module is covered until its adapter is implemented, tested against a fixture, and listed in the export profile registry with a coverage percentage. Expanded coverage will land in the registry first; the roadmap page will reflect the registry numbers automatically.

New buyer-ask templates are added when a supplier in the cohort needs one. Each new template is loaded with the same source-traced autofill discipline as CDP Climate, EcoVadis, and GRI Universal — quantitative pre-fill from evidence, qualitative prompting, no silent self-declaration.

The pinned factor library is DEFRA 2025, UK only. Non-UK suppliers fall back to the UK factor today — a known limitation called out in the methodology page. Multi-geography factor sets are on the roadmap; they ship when we can pin a specific edition per geography and stand behind reproducibility the same way we do for DEFRA today.

Honest limits

Everything above is intent, not commitment. We do not publish dates for any of these items. We will not move a Partial item to Live status until it is shipped, tested, and the registry numbers reflect reality. If an item drops off this list in a future revision, that revision will say so explicitly.

A roadmap is also a list of what we deliberately are not doing. The items below are not delays, slips, or pending decisions — they are deliberate scope exclusions. Each carries a brief why-not so an evaluator can judge whether our reasoning lines up with their needs.

The EU Corporate Sustainability Reporting Directive requires submission in inline XBRL (iXBRL) against the ESRS taxonomy. That is a substantial architectural commitment — full taxonomy mapping, structured tagging, validation against the official EU schema — and it is not near-term. Suppliers needing CSRD-grade filing today should treat TSK as a complementary evidence base, not a filing tool.

TSK is supplier-first evidence-extraction software. Each tenant's data stays in their tenancy. We do not centralise supplier evidence across customers, we do not act as a clearinghouse between buyers and suppliers, and we do not run a shared data network. The product definition is locked on this — it is a different category of product, not a stretch goal.

Standards bodies — CDP, GRI, ISSB, DEFRA — publish methodology updates on annual or off-cycle schedules. TSK does not subscribe to those publications, does not auto-update factor sets, and does not auto-update template scoring rules. Updates land via versioned release cycles after manual verification of each change. That trade favours reproducibility over freshness, deliberately.

Honest limits

If any of these positions changes, it will move out of this section first and into the Next quarter section — never the other way around. A non-goal becoming a goal is a meaningful product decision that we will document with reasoning, not slide quietly in or out of scope.