Training · Evaluation · Retrieval

Business data, sourced for your AI task.

Maevar helps AI labs and product teams source relevant records from specialist small and medium-sized businesses. Emails, quotations, engineering files, delivery records and production histories can connect the initial request to the decisions, exceptions and outcomes that follow.

Tell us what you want a model to do, the sectors that matter and which records need to connect. We turn that requirement into a sourcing brief, investigate suitable businesses and records, and establish the fit, rights and next steps with you.

Discuss your use case Explore the data

The examples below describe sourcing opportunities. Availability, licensing authority and any privacy or confidentiality restrictions need to be established for each project.

Context beyond a single document

A document shows one moment. A workflow shows what happened next.

A quotation may show a price, but not why it changed. A delivery note may confirm a shipment, but not the missed collection, revised instruction or customer complaint that preceded it. A machine log may record a stoppage without identifying the job, maintenance action or quality issue around it.

For many tasks, the useful starting point is a linked sequence: what was requested, what information was available, what changed, what someone decided and what outcome was recorded. Maevar starts by defining that sequence with you, then investigates whether suitable records and permissions can be assembled around it.

Find the right evidence

Retrieve the relevant email, drawing revision, order or procedure from a mixed set of records, and cite the source used.

Follow a change through the work

Trace an amended specification through a revised quotation, purchase order, production instruction and final invoice.

Understand exceptions

Examine a late delivery, rejected part, returned product or disputed charge using the records available at the time.

Recognise what is missing

Distinguish a supported conclusion from an assumption, an unresolved discrepancy or a question that needs human review.

Start with the evidence you need

Business records, in useful detail.

A brief can focus on one record type or a connected set across several systems. The following families show the kinds of material that could be investigated, the relationships that may matter and the AI tasks an evaluation could test.

Linked workflow and case histories

Link an enquiry, its correspondence, job records and close-out notes to reconstruct what happened, identify an outstanding action or explain an exception.

Records, connections and AI tasks

Records to explore

Initial requests, task or job records, correspondence, attachments, status changes, approvals, exception notes, handovers and close-out records. A case may span an inbox, spreadsheet, operational system and document folder.

Connections that matter

Case identifiers, timestamps, document versions, the order of events and links between the request, action and result. A useful history may include failed attempts and unresolved issues as well as completed work.

Tasks to scope

Reconstruct a timeline; identify the next action supported by the current case; summarise a handover; explain a delay; compare a requested outcome with the recorded result.

Assess before use

How complete the history is, whether timestamps are reliable, which links are explicit and which have been inferred, and whether later information must be withheld when testing an earlier decision.

Email, internal communication and CRM

Connect emails, CRM notes and support threads to the relevant quote, order or issue, so an answer can recover the customer’s requirements and the commitments still open.

Records, connections and AI tasks

Records to explore

Email threads and attachments, internal messages, CRM notes, lead and opportunity histories, customer enquiries, support tickets, service notes and recorded follow-up actions.

Connections that matter

Link a conversation to the relevant customer case, quotation, order, project or issue. Preserve thread order, attachment references and changes in status where the source allows it.

Tasks to scope

Retrieve an answer with evidence; extract a customer's requirements; identify outstanding commitments; classify an issue; summarise a long case; draft a response grounded in the authorised records.

Assess before use

Personal information, confidential negotiations, third-party correspondence and whether the business has authority for the intended use. A mailbox export does not establish permission to license everything inside it.

Quotations, pricing and commercial changes

Connect specifications, quote revisions, supplier prices and accepted orders to explain a commercial change or identify which scope and price were agreed.

Records, connections and AI tasks

Records to explore

Enquiries, specifications, estimates, quotation versions, line-item prices, supplier price lists, proposals, scope changes, customer questions, accepted orders and relevant contract documents.

Connections that matter

Match revisions to the original request, identify changed quantities or specifications, and connect the accepted offer to the order and subsequent fulfilment or project record.

Tasks to scope

Compare quotations; extract inclusions and exclusions; explain a price change; flag inconsistent quantities; identify the accepted revision; produce an evidence-backed scope summary.

Assess before use

Whether quoted and accepted prices are distinguishable, how discounts and tax are represented, whether supplier information can be shared and whether the acceptance record is present.

Invoices, credits and financial administration

Link orders, invoices, credits and delivery or completion records to reconcile a charge and distinguish what was billed, adjusted or recorded as paid.

Records, connections and AI tasks

Records to explore

Purchase orders, sales and supplier invoices, credit notes, delivery or completion records, remittance information where appropriate, reconciliation notes and correspondence about disputed or adjusted charges.

Connections that matter

Connect the ordered item or service to receipt or completion, invoicing, corrections and any recorded payment status. Keep quoted, invoiced, credited and paid amounts distinct.

Tasks to scope

Match documents; extract line items; reconcile totals; explain a credit; identify a missing reference; flag a discrepancy for review without inventing its cause.

Assess before use

Sensitive financial details, bank information and identifiers; consistent currencies and tax treatment; and whether the records actually establish payment or only an amount billed.

Procurement, stock and fulfilment

Connect purchase orders, stock movements, dispatch records and proof of delivery to trace a shortage, partial shipment, substitution or unresolved return.

Records, connections and AI tasks

Records to explore

Purchase requests, supplier quotations, purchase orders, stock records, goods-received notes, allocation and picking records, dispatch notes, manifests, proof of delivery, shortages, substitutions and returns.

Connections that matter

Product or part identifiers, order lines, locations, batches where recorded, quantities and event dates. Link an exception to the original requirement and the actions taken to resolve it.

Tasks to scope

Reconcile ordered and received quantities; trace a shortage; compare promised and recorded delivery dates; identify an unresolved return; answer a stock or order-status question from the available evidence.

Assess before use

Inconsistent item codes, partial deliveries, stock snapshots versus transaction histories, missing events and commercially sensitive supplier or customer information.

Product catalogues, ecommerce and support

Link product variants and catalogue versions to orders, shipments and support cases to answer specification questions or investigate a fulfilment or return issue.

Records, connections and AI tasks

Records to explore

Product descriptions, specifications, variants, catalogue revisions, stock and order exports, fulfilment events, return reasons, support tickets, FAQs and authorised product documentation.

Connections that matter

Join product and variant identifiers to orders, shipment records, returns and support cases. Keep the product information that applied at the time of the transaction where it can be established.

Tasks to scope

Answer specification questions; compare variants; extract product attributes; assess whether a support response uses the right product evidence; explain a return or fulfilment exception.

Assess before use

Rights to manufacturer content and product images, personal customer data, changing catalogue information and whether outcomes such as refund or replacement are explicitly recorded.

Engineering, CAD and CNC job records

Connect drawing revisions, CAD files, CNC programmes, job cards and inspection records to check which requirement applied and how the recorded work relates to it.

Records, connections and AI tasks

Records to explore

Engineering drawings, CAD files or exports, revision histories, specifications, bills of materials, CNC programmes, setup sheets, tooling notes, job cards, inspection reports and non-conformance records.

Connections that matter

Link the part and drawing revision to the quotation, production job, programme or setup and inspection outcome. The relationship between files can be as important as the files themselves.

Tasks to scope

Retrieve the applicable revision; compare requirements; identify a mismatch between a drawing and job instruction; connect an inspection finding to its source specification; summarise a technical change.

Assess before use

Customer and third-party intellectual property, file interpretability, units and revision control, and the availability of appropriate domain review. Historical programmes or instructions are not automatically safe or correct for execution.

Production, quality, maintenance and repair

Join production runs, machine and maintenance logs, quality checks and rework records to reconstruct a stoppage or investigate a recorded output or quality issue.

Records, connections and AI tasks

Records to explore

Production schedules, work orders, batch records, machine logs or sensor exports where recorded, downtime entries, material usage, scrap or rework records, quality checks, fault reports, diagnostic notes, parts selection, service reports and corrective-action notes.

Connections that matter

Relate events to the machine, job, product, batch and time period where identifiers exist. Link a reported problem to the inspection, repair or follow-up outcome rather than treating each log entry in isolation.

Tasks to scope

Classify downtime; reconstruct a maintenance case; find repeat recorded issues; explain an output discrepancy; retrieve a relevant procedure; summarise the evidence behind a quality investigation.

Assess before use

Log resolution, timestamp alignment, missing operating context, consistency of fault labels and whether a reported cause was verified or merely suspected. Predictive uses would need a separate assessment of coverage and labels.

Procedures, contracts and project decisions

Connect SOPs, contract terms, checklists and decision logs to the relevant task and date, so a model can retrieve the applicable instruction and identify unresolved obligations.

Records, connections and AI tasks

Records to explore

SOPs, work instructions, checklists, handover notes, meeting records, project decision logs, agreed specifications and contracts or amendments that the source is authorised to share.

Connections that matter

Associate the applicable document version with the task or decision. Link a change in instructions to its stated reason and any evidence of implementation.

Tasks to scope

Retrieve the relevant procedure; compare versions; extract obligations or deadlines for human review; identify an unresolved action; test whether an answer distinguishes a stated rule from a practice observed in the records.

Assess before use

Confidentiality, ownership and third-party rights; whether a procedure was current at the relevant time; and the distinction between a recorded contractual term, an interpretation and legal advice.

Design, marketing and creative work

Connect briefs, working files, creative revisions, approvals and campaign records to understand what was requested, what changed and what was eventually released.

Records, connections and AI tasks

Records to explore

Design or campaign briefs, authorised working files and assets, copy and artwork versions, feedback, approval records, channel plans, advertising histories and associated performance reports.

Connections that matter

Link a brief and its requirements to the relevant asset version, review comments, approval and publication or campaign record. Keep reported outcomes separate from any explanation of what caused them.

Tasks to scope

Compare the final work with the brief; trace the reason for a revision; identify the approved asset; summarise outstanding feedback; retrieve evidence for a campaign review.

Assess before use

Rights to creative assets, stock material, brands, likenesses and third-party content; client confidentiality; and whether performance records can be reliably linked to the work being assessed.

Have a particular record type in mind? Tell us the task and the connections you need. →

Different sectors. Specific evidence.

What a sourcing brief could look like.

Start with a question and work back to the evidence. These illustrative briefs show the records, relationships and assessment points that matter in different sectors.

Haulage and logistics

Booking → dispatch plan → manifest → delivery event → exception → resolution.

Potential records

Transport bookings, dispatch sheets, routes and schedules, load manifests, driver or depot updates, proof of delivery, failed-delivery notes, customer communications and related charges.

Example task

“Why was this delivery late, what did the customer get told and which parts of the explanation are documented?”

What to establish

Reliable links between the booking and delivery, the sequence of revised instructions, whether the recorded route reflects a plan or an actual journey, and the treatment of driver and customer information.

CNC and precision engineering

Enquiry → drawing revision → quote → job setup → manufacture → inspection.

Potential records

Customer specifications, drawings and CAD exports, quotation revisions, material and tooling requirements, CNC programmes, setup sheets, job cards, inspection results and rework notes.

Example task

“Which drawing revision applied to this job, and does the inspection record address its stated tolerances?”

What to establish

Revision consistency, permissions for customer-owned designs, meaningful links between programmes and jobs, and the domain expertise required to assess an answer.

Manufacturing and plastics

Production order → materials and setup → run → quality checks → rework or release.

Potential records

Bills of materials, material grades and lots, mould or tool identifiers, batch and run sheets, recorded process settings, production counts, scrap, downtime, maintenance and quality investigations.

Example task

“What changed between these production runs, and which recorded facts support the explanation for additional scrap?”

What to establish

Whether the material lot, mould or tool, machine and run can be matched across systems; whether measurements are comparable; and whether the explanation is confirmed, disputed or absent.

Printing and packaging

Customer specification → estimate → artwork or proof → approval → print job → finishing and dispatch.

Potential records

Job specifications, quotation changes, artwork versions where rights permit, proof approvals, job tickets, substrate and finishing instructions, production checks, reprint records and delivery notes.

Example task

“Did the final job reflect the approved specification, and where did the reprint request originate?”

What to establish

The approved version, the sequence of changes, rights to artwork and customer content, and whether a defect or reprint reason was documented.

Wholesale, ecommerce and product businesses

Product listing → customer order → stock allocation → fulfilment → support or return.

Potential records

Catalogue and variant data, supplier quotations, stock histories, substitutions, orders, dispatch events, support threads, return reasons, authorised manuals and recorded resolutions.

Example task

“Was the customer's question answered using the right product variant, and what happened after the proposed resolution?”

What to establish

Stable product identifiers, the information available at the time, customer-data restrictions and a recorded link between the support case and its outcome.

Construction and project delivery

Enquiry and scope → estimate → instruction or variation → valuation → final account.

Potential records

Estimates, bills or schedules of quantities, quotations, tender clarifications, project correspondence, instructions, variation records, progress records, valuations and final accounts.

Example task

“Why did the final cost change, which changes are documented and what remains unexplained?”

What to establish

Consistent scope and tax treatment, links between an instruction and its commercial effect, evidence of approval, and the difference between an amount claimed, agreed, invoiced or paid.

Explore the synthetic construction example →

Maintenance, repair and field service

Fault report → diagnosis → parts or repair decision → service work → test and outcome.

Potential records

Customer fault descriptions, diagnostic histories, equipment details, parts selections, repair specifications, technician notes, service records, follow-up questions and test results.

Example task

“What evidence supported the repair choice, which checks were performed and was the reported fault resolved?”

What to establish

Equipment and job identifiers, the sequence of tests and interventions, whether a diagnosis was confirmed and whether the outcome was checked rather than assumed.

Business services and commercial operations

Enquiry → proposal → agreement → delivery → handover → invoice or renewal.

Potential records

CRM histories, proposals, service specifications, authorised contracts, delivery correspondence, project plans, action logs, handovers, invoices and support records.

Example task

“What was agreed, what changed during delivery and which commitments are still open?”

What to establish

A clear case boundary, the accepted agreement or scope, changes in responsibility and permission to use client communications.

Beyond text and spreadsheets

Where a brief needs images, audio or demonstrations.

Some research questions need a visual or time-based record of the work. We can discuss whether these formats belong in a sourcing brief and what would have to be established before a source could be considered.

Images and video

Potential examples include authorised product or inspection images, recorded process demonstrations, first-person footage and workplace or CCTV recordings. The task might involve recognising a documented state, relating a visual observation to a work record or following a sequence of actions.

Audio and conversations

Potential examples include recorded service conversations, spoken instructions or consented demonstrations. A brief would need to specify whether it needs the recording, a transcript, speaker or timing information, or links to an associated case. The capture context and any alignment with video or operational records would need to be established.

Software demonstrations and screen recordings

Potential examples include an authorised demonstration of completing a business task in software, with the visible steps connected to a task description and outcome. Credentials, personal information, customer records and restricted software content would need specific consideration.

Robotics-related product and task material

Potential examples include authorised robot product documentation, manuals, specifications, capability or compatibility information and demonstrations. A product catalogue is different from a robot action dataset. Requests for sensor streams, trajectories, synchronised observations or annotated physical actions would require a separate technical and sourcing assessment.

These are exploratory formats, not a claim that Maevar holds, has licensed or can currently deliver these materials. Rights to recording, people depicted or heard, locations, equipment, software and underlying content would need review. Removing names alone may not resolve the risks.

One sector, one worked example

Why did the final cost change?

This synthetic construction case makes the sourcing idea concrete: connect the records, reconcile the figures, cite the evidence and say what cannot be established. A different sector would require its own records, task definition and assessment criteria.

SCRIPTED SYNTHETIC DEMONSTRATION · All records and figures are invented. This is not a live AI system or a real archive.

Where did the extra £12,000 go?

Select a record, then follow the citations. Remove the variation summary to see how the evidence changes.

V1 / Variation summary

Line 1 Additional drainage: +£12,000, excluding VAT.

Line 2 Omitted partitions: −£2,000, excluding VAT.

No supporting approval or invoice is included.

Include variation summary

£12,000 difference between estimate and final account

£10,000 documented changes£2,000 unexplained

What the records support

The recorded final cost is £12,000 above the estimate (F1, line 1; E1, line 1).

The variation summary documents £10,000 of that difference: £12,000 for drainage (V1, line 1) less £2,000 for omitted partitions (V1, line 2).

£2,000 remains unexplained. An itemised final-account reconciliation is needed.

All amounts exclude VAT. The records do not establish whether variations were approved or amounts paid.

Discuss a similar task →

This interactive example needs JavaScript. With V1 included: £100,000 estimate + £10,000 documented changes + £2,000 unexplained = £112,000 final. Without V1, the entire £12,000 increase is unexplained.

What a real evaluation could test

Whether the answer reconciles numbers across separate records.
Whether each factual claim can be traced to a source line.
Whether the answer changes when evidence is removed.
Whether missing approval, payment or reconciliation evidence stays explicitly unresolved.

A real evaluation would need an agreed source set, task design and review method. This scripted example demonstrates the questions to ask; it does not demonstrate the performance of a model or the quality of a real dataset.

Agree what useful means

Define the dataset around the task.

A file count alone does not say whether a source will help your team. Before committing to preparation or delivery, the discussion should establish the unit of work, the evidence it contains, the intended use and how suitability will be judged.

Training and adaptation

Define the subject matter, source mix, coverage and intended training use. Consider how repeated material, uneven coverage and later preparation affect the corpus. The licence must cover the planned use; the presence of records alone does not establish training suitability.

Evaluation

Define coherent cases, the inputs the model can see, any outcomes to hold back and the criteria for assessing its response. Tasks can focus on extraction, reconciliation, grounded answers or decisions within a workflow, with uncertainty kept visible.

Retrieval and evidence-grounded tools

Define which source versions a system should retrieve, what access restrictions apply and how answers should cite supporting records. Keeping an answer connected to its evidence is a different requirement from creating a training corpus.

Scope and coverage

Which sectors, business processes, date ranges and record types are relevant? Is the useful unit a document, conversation, order, production run or complete case? Which routine and exceptional cases need to be represented?

Structure and relationships

Which source formats are available? Which fields, identifiers, timestamps and revision links can be preserved? Where extraction or conversion is needed, what fidelity checks would be required? Original files, tables or structured records are options to assess, not a fixed delivery promise.

Provenance and gaps

What can be documented about the source and preparation? Which records are missing, duplicated, inconsistent or incomplete? Distinguish original source facts, observed historical outcomes and later annotations. Document the method used to link records and flag uncertain or inferred relationships.

Task design and assessment

What will the model be asked to do, what evidence can it see and how will an answer be assessed? If reference answers, labels or specialist review are needed, agree the method, review requirements and handling of disagreement before treating them as a deliverable.

Evaluation boundaries

Define sample size, access, permitted uses, retention, onward sharing and any restrictions. Decide whether separation by business, job, case or time period is needed to keep related records or later outcomes from influencing the test.

Licence and delivery

Agree the parties, material covered, authorised uses, restrictions, commercial terms and delivery method. Training, evaluation, retrieval, derived outputs, redistribution, retention and any exclusivity should be addressed explicitly where relevant. A feasible source is not a licence until the necessary rights and terms are agreed.

A delivery package to agree

Depending on the source and project, the scope can specify:

The agreed source records or extracts, with their intended format and level of fidelity.
A record manifest identifying cases, source files, dates and versions where available.
A relationship map showing how records connect and which joins are explicit or inferred.
Preparation notes covering conversions, exclusions, quality issues and missing evidence.
A summary of the permitted uses and relevant restrictions, alongside the agreed licence.
Labels, reference answers or specialist review only where separately scoped and agreed.

These are specification options. The source assessment establishes which can be produced, by whom and at what cost.

What to check in a proposed sample

The unit being assessed: a file, a conversation, a linked case, a job or a production run.
Case counts as well as file counts, date range, business and source mix, and balance of routine and exceptional cases.
Whether the required files and fields are readable and the essential record links are present.
Which joins are source-backed, inferred, uncertain or missing, and what that means for the task.
Whether answers can be assessed against the permitted evidence without revealing later outcomes.
The rights, restrictions and preparation still needed before any wider use.

The acceptance criteria should be agreed before the sample is treated as suitable. A collection of records is not automatically an evaluation environment; any task interface, annotation or scoring work needs its own scope.

The outputs of scoping may be a feasible evaluation, a narrower brief or a clear finding that the proposed source cannot support the task.

From brief to agreed use

Start with the task. Establish what is feasible.

01. Define the buyer brief
Tell us what your team wants to test or build, the records it expects to need and the context that makes an example useful. Include relevant sectors, formats, date ranges, scale and timing if you know them. A short non-confidential description is enough to begin.
02. Investigate sources and fit
Assess potential businesses and records against that brief. Look at availability, structure, links, completeness and the work needed to make a sample assessable. Raise gaps early rather than treating the existence of an archive as evidence that it fits the task.
03. Assess rights, privacy and boundaries
Establish who can authorise the intended use and what restrictions apply. Consider confidentiality, third-party content and personal information. Define exclusions and any preparation or access controls before material is shared.
04. Scope a bounded evaluation
Where a suitable source and permissions can be established, agree a sample or evaluation: its content, permitted purpose, preparation, access, review criteria and commercial basis. Use that scope to decide whether further work is justified.
05. Agree the licence and delivery
If the project proceeds, agree the material, permitted uses, preparation, delivery method and responsibilities. Any further sourcing or expansion is assessed against the agreed requirements rather than assumed to be included.

Source availability, preparation, specialist review, timing and commercial terms are established project by project.

A founder-led sourcing initiative

Practical business context. A buyer-led brief.

Maevar is led by Richard McKay, who holds an honours degree in Quantity Surveying and has worked in the construction industry for over 30 years.

His construction background informs that sector. Sourcing in haulage, engineering, manufacturing, commerce and other fields starts with the relevant business, the records it holds and the expertise needed to understand them.

Maevar is an early-stage business data-sourcing initiative. The aim is to connect a clearly defined AI task with appropriate business evidence and a workable route to permitted use. Any specialist review, preparation and evaluation requirements would be agreed for the project.

For business owners and authorised archive holders: discuss potential records with McKays.

Useful to know

Questions before a brief.

Are these datasets available to buy now?

The categories on this site describe potential sourcing briefs. They are not a stock list, and no specific availability, volume or delivery date is implied. We start with your requirement and establish whether suitable sources, permissions and preparation can be put in place.

Do you work only with construction data?

No. Construction is the founder's professional background and the subject of the synthetic example. The sourcing brief can cover other sectors, including haulage, precision engineering, manufacturing, printing, ecommerce and business services. Feasibility and any necessary sector expertise are assessed for each project.

What should we include in an initial brief?

Describe the task, who will use the data and what a useful example looks like. Add the sectors, record types, relationships, date range, expected scale, formats and intended use where known. Explain whether you need individual documents or complete cases, and what would make a sample unsuitable. Please keep the first enquiry non-confidential.

Can we start with a small evaluation?

A bounded evaluation can be discussed where a suitable source and the necessary permissions can be established. Sample content, access, preparation, permitted use, retention and any charges would be agreed before records are shared. A sample is not automatically licensed for later training or redistribution.

Can the data be used for model training?

Potential use depends on the source and the licence that can be agreed. Training, fine-tuning, evaluation, retrieval, derived outputs and onward distribution are different uses and should be specified in the brief. Retention and any exclusivity also need agreement. Access to records does not itself establish rights to use them for any of these purposes.

How are personal and confidential records handled?

The source needs assessment for personal information, confidentiality obligations, third-party content and authority to license. The outcome may require exclusions, redaction, restricted access or a different brief. Some records may be unsuitable. We do not assume that removing names makes a dataset safe to share or use.

Can you provide task banks, reference answers or evaluation environments?

These can be requirements to explore in the project scope, alongside labels and expert review. A task bank or interactive environment requires work beyond supplying records: task design, interfaces, scoring and appropriate expertise. Availability, methods, quality checks and cost would need to be established separately. Labels created later should be distinguishable from source facts, and disagreement or uncertainty should not be concealed.

Are historical business decisions treated as correct answers?

No. A history records what someone did, not necessarily what they should have done. A useful evaluation may test factual extraction or evidence tracing without endorsing the original decision. Judgement-based tasks need an agreed review and scoring method.

What formats, volumes and timescales can you support?

Those depend on the source and the preparation required. Tell us your preferred formats, useful sample size, larger-scale requirement and deadline. Feasibility, fidelity, delivery method, volume and timing should be confirmed during scoping rather than inferred from the examples on this site.

Can you source video, audio or robotics data?

These can be discussed as exploratory sourcing requirements. They involve different technical, rights and privacy questions from ordinary business documents. Maevar is not claiming current inventory or delivery capability for these formats. A specific brief is needed to establish whether a suitable source and lawful route to use exist.

How is pricing agreed?

Pricing would depend on the source, permitted use, preparation, review requirements, volume and delivery arrangements. Any evaluation fee, licence fee or further work would be agreed in the project scope. The site does not quote a standard per-record price or imply that an evaluation is free.

Is the worked example a real archive or live AI system?

No. Its records and figures are invented and its answers are scripted. It illustrates a task involving linked evidence and missing information. It is not evidence of dataset inventory, customer work, model performance or access to a real business archive.

Tell us what your team needs

Start with the task.

You do not need a finished specification. Tell us what you want a model to do, which records might support it and where the evidence is currently missing. We will review the requirement and reply about fit and possible next steps.

For business owners and authorised archive holders: discuss potential records with McKays.