check_record
ChatGPTCheck the CAST Record — a curated index of entertainment-technology companies (venues, live events, touring, broadcast, music tech) — for companies matching a challenge, need or capability. Use whenever someone says 'check Record', 'check the Record', 'ask Record', or wants entertainment/venue tech vendors. Present the answer as given under the heading 'Record results', each company linked to its record_url so the reader can open the full profile, and add nothing of your own; the result's present field says how. A question about what is changing, where an area is heading or trends is answered by depth researched (its brief says where the field is moving, from Record's dated readings) together with market_landscape (the counts), never by quick, which returns companies only. quick returns ranked matches in a few seconds; researched adds a synthesised answer and per-company reasoning but takes about half a minute, longer for a plan. If the person has shared a strategic plan, board paper, tender or brief with you, pass it: either as org_context, or by including the relevant part in the question itself. Record then ranks for fit against it and returns brief — the plan assessed against where the field is moving, priority by priority, with the companies last — and brief_read, the working behind it.
comparable_awards
ChatGPTWhat comparable buyers ACTUALLY PAID, from published public-sector award notices. Use it whenever price, budget or value for money comes up — 'what should we budget for access control', 'is 40k reasonable for a ticketing platform', 'what do venues pay for this'. Record declines to make a pricing judgement and that does not change; this hands over the evidence underneath it instead. Present the MEDIAN and the range, never a mean — procurement values are wildly skewed and one framework covering forty authorities sits in the same set as a single venue's purchase. Quote every figure with its buyer, date and notice link, and say plainly that scope differs between notices (licences alone versus licences plus five years of support), so this is a range to reason from and not a price for the reader's own situation. Reached by 'ask Record what venues pay for this' as much as by a bare question. Present as given; add nothing of your own. Fast (~1s).
get_company
ChatGPTFetch the full CAST Record profile for one company by its Record slug (as returned by check_record). Present it as given; add nothing of your own. Fast (~1s).
index_stats
ChatGPTCoverage and freshness statistics for the CAST Record index (recent changes, update cadence). No input. This is also where 'ask Record what it covers' belongs. Present as given; add nothing of your own. Fast (~1s).
lookup_company
ChatGPTCheck whether a specific company is already in the CAST Record index, by website domain (preferred) or name. Use this for 'is X on Record?' questions rather than searching, including when it is put as 'ask Record whether X is listed'. Fast (~1s).
market_landscape
ChatGPTCOUNTS, not a search. Use this — never check_record — whenever the question is about the shape of the field rather than about which companies to shortlist: where the opportunity is, what is underbuilt, which categories are growing or converging, what is being built in one region versus another, which problems have the most companies chasing them, what people are asking for and not finding. check_record answers those with a similarity search, which cannot count and will present ten retrieved rows as a finding about the industry. This returns arithmetic over the whole index instead: companies by category, coverage growth, stated region with the unstated count beside it, founding years, which PROBLEM DOMAINS the reporting is concentrating in (news.problems — the axis that aggregates, and the one to read for where the field is under pressure; news.themes is per-story prose and is filtered to labels carried by more than one reading, so it is often short), what it connects, and — the one measure of DEMAND rather than supply — the needs that reached Record and found nothing. Every figure carries its window and its population. Read the reading field before quoting any of them, and follow it: quote the counts as given and add no read of your own. Never call this to add to an answer another tool already gave, a brief above all; only when the reader asks about the field itself. Reached by 'ask Record where the field is heading' as much as by a bare question. Takes 3-6s.
works_with
ChatGPTThe REVERSE lookup, and not a search. Use it when the reader already has a system they cannot replace, or names an organisation and wants to know who serves it: 'what works alongside Tessitura', 'which vendors integrate with Spektrix', 'which suppliers name the OVO Hydro as a customer'. check_record cannot answer these — it finds companies that DO a thing, not companies that NAME a thing. Both lists here come from names each company published on its own integrations or customers page, matched exactly. Absence is never a denial: say a company 'does not name it' rather than 'does not support it', because many publish no list at all. names_as_customer is a published logo and never private intelligence about what an organisation runs. named_by_venues, when present, is the operating stack around it at the venues themselves: of the venues Record has read that name it on their own pages, how many also name each other supplier, given only from 25 such venues; quote every count with its base and date. Reached by 'ask Record what works with X' as much as by a bare question. Present as given; add nothing of your own. Fast (~1s).
