Skip to main content

AI-01 AI integration · Worldwide

AI Integration Services

A demonstration proves a model can do something once. We measure it on your own documents, tell you the accuracy per field, set the threshold below which a person checks it, and put in writing where your data goes.

  • A pilot on 50 to 200 real examples, with a measured accuracy figure
  • Accuracy reported per field, because one average hides the problem
  • A human review threshold, so low confidence reaches a person
  • Provider, region, training and retention answered in writing

Reply within 1 working day · the test set stays yours

P J DivyaCo-Founder, CTO and CMO

Builds the retrieval layer, runs the evaluation, and sets the review threshold with you.

What a pilot returns

Measured Accuracy per fieldConfusion breakdownCost per document
Designed Review thresholdSchema validationFallback path
Documented ProviderRegionTrainingRetention
Monitored DriftReview rateFailure rate
  1. InputOne use case worth measuring
  2. GroundYour data as the source
  3. EvaluateAccuracy measured, not claimed
  4. OutputA human review threshold
StartMeasured pilot
Pilot2 to 4 weeks
OutputA measured number
DesignHuman in the loop

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

  • 01

    Extraction

    Pulling fields out of invoices, purchase orders, bank statements, contracts and forms that arrive as scans and PDFs in no fixed layout.

    Highest value
  • 02

    Classification

    Sorting email, tickets, complaints and documents into categories so they route to the right queue without a person reading each one.

    Highest volume
  • 03

    Summarisation

    Reducing long threads, call transcripts, reports and agreements to the points a specific reader needs, with the source retained.

    Time saver
  • 04

    Drafting

    A first version of a quote, a reply, a job description or a report, which a person then edits. Faster to correct than to start.

    Assisted
  • 05

    Search over your own content

    Asking a question and getting an answer from your own documents, with the passage it came from shown so it can be checked.

    Retrieval
  • 06

    Translation

    Moving content between the languages your market actually reads at a quality worth editing, which is a different claim from publishing it unread.

    Assisted

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

Four tasks we decline, and what we suggest instead
The askWhy it failsWhat we suggest
Arithmetic that must reconcileA model produces plausible numbers, and plausible does not tie out to the paisaCompute in code, use the model only to find the inputs
A decision with legal consequenceCredit, hiring and statutory positions need an accountable humanModel prepares the file, a person decides and signs
A figure going straight into a returnAn unreviewed output in a filing is a defect with a deadline attachedExtraction plus mandatory review before submission
A task nobody can markIf no one can say whether an output was right, accuracy cannot be measuredDefine the correct answer first, or do not automate it

On the word hallucination

Every model sometimes produces a confident answer the source does not support. Retrieval with citations, a confidence threshold, schema validation and a review queue reduce it to a manageable rate. Any vendor telling you their system does not hallucinate is describing a demo rather than a deployment.

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

  1. Step 1: Define the correct answer

    For each example, what your team would say is right. If two of your own people disagree, that disagreement is the first finding and it matters more than the model.

    Days 1 to 3
  2. Step 2: Build the test set

    Between 50 and 200 real documents or messages, covering the awkward cases rather than the tidy ones. A test set of clean examples measures nothing useful.

    Days 3 to 7
  3. Step 3: Run and count

    Accuracy per field, not overall. An invoice extractor at 95% overall can be 99% on the amount and 70% on the vendor name, and only the per-field view shows it.

    Week 2
  4. Step 4: Read the failures

    What kind of document it fails on, and whether prompting, retrieval or a validation rule fixes it. Most of the improvement comes from here rather than from a bigger model.

    Week 2 to 3
  5. Step 5: Recommend, including not building

    A written view with the accuracy, the cost per document and the review load it implies. We have recommended against building often enough for that to be a real outcome.

    Week 3 to 4

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

The same model, three different thresholds
ThresholdGoes straight throughReaches a personErrors that escape
None100%0%All of them, into your process
Low confidence only92%8%The rare confident mistake
Conservative70%30%Almost none, at a real review cost
  • The threshold is set from what an error costs, not from what looks good in a report. A misread invoice amount and a misrouted support ticket do not deserve the same setting.
  • Everything below the threshold goes to a queue with a named owner and a service level, exactly as an approval workflow would.
  • The review rate is monitored, because a rising share below the threshold is the earliest signal that the incoming documents have changed or the provider has updated the model.
  • Every output is validated against a schema before it is accepted at all, so a malformed answer fails loudly rather than entering the process quietly.

The useful part was being told the vendor name field was only 70% accurate. We set that one to always review and let the amounts through, and the process worked from week one.

Finance managerLogistics company

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

  • Which provider, and which region. Named, with the processing location stated. Where data must remain in a named jurisdiction, that constrains the model choice and we say which options are then available.
  • Whether your inputs are used for training. Answered per provider and per plan, because the default differs between consumer and business terms and the difference matters.
  • What retention applies. How long the provider holds a request, and what we hold in logs. Both are configurable and both are decided before launch rather than discovered later.
  • What personal data is involved. Sending personal data to a model is processing it, so the General Data Protection Regulation and its equivalents apply. We map the fields and remove what the task does not need.
  • What never leaves. Often more than clients expect. Redaction before the request, retrieval over an index you host, and processing only the extract rather than the whole document are all ways of shrinking what is sent at all.

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

The evaluation set is built before any model is chosen, and the people who build it are the people who report the numbers.

  • P J Divya

    Co-Founder, CTO and CMO

    Grounds the model in your data, wires it into the system of record, and scores it against a labelled set.

  • Chetan Ravi

    Founder and CEO

    Picks the one use case worth measuring and decides what accuracy has to be worth in money.

  • Harith, Compliance Expert at ProTax

    Harith

    Compliance Expert

    Checks retention, residency and what the model may be shown from your records.

  • Harsh Jain, Legal Expert at ProTax

    Harsh Jain

    Legal Expert

    Answers in writing where your data goes, who processes it, and what the provider may train on.

10 Related

Where AI gets applied

This page is the method and the governance. These are the two places it gets used most.

  • Working hours

    Our day runs on UTC+5:30. The overlap window with your team is written into the scope rather than assumed, and everything outside it runs asynchronously.

  • How we communicate

    One written update a day on the channel you already use, a standing weekly call inside the overlap window, and a named person to escalate to. Nothing important is agreed only on a call.

  • Who you contract with

    ProTax Consultancy, the company behind ProTax, named in the agreement with its registration number. The governing law and the forum are agreed before you sign, not after a dispute.

  • Currency and payment

    Invoiced in your currency or in ours, your choice, and settled by bank transfer. Milestones are tied to deliverables you can see, never to elapsed time.

  • What you own

    Copyright in everything built for you is assigned on final payment: source code, design files, prompts, configuration and documentation. Third-party licences are listed by name so nothing is a surprise later.

  • Where data sits

    You choose the region your data is stored and processed in, and the answer is written down before the build starts, with who at ProTax can reach it and for how long.

Offset from our working day standard time

  • London-5:30
  • Dubai-1:30
  • Singapore+2:30
  • Sydney+4:30
  • New York-10:30
  • San Francisco-13:30

ProTax is a brand of ProTax Consultancy. The contracting entity, the governing law and the invoicing currency are all named in the proposal before you sign anything.

09 Questions

Where AI gets applied

Large language model
Written LLM: a model that produces plausible text continuations. Excellent at drafting and extraction, never a source of guaranteed correctness.
Hallucination
A confident output the source material does not support. Reduced by retrieval, thresholds and validation, and never eliminated.
Retrieval augmented generation
Written RAG: answering from your own indexed documents, with the passage shown so the answer can be checked.
Test set
A fixed collection of real examples with known correct answers, used to measure accuracy before and after any change.
Human review threshold
The confidence level below which output is routed to a person rather than accepted automatically.
Token
The unit a model provider bills by, roughly a fragment of a word. It is why AI cost scales with volume rather than with user count.
Schema validation
Checking an output has the expected shape and types before it is accepted, so a malformed answer fails loudly.
GDPR
The EU and UK data protection regime. Article 83 sets administrative fines of up to €20 million or 4% of worldwide annual turnover, whichever is higher, which is why consent, retention and access control are build decisions here rather than paperwork.

09 Questions

Where AI gets applied

Connecting a language or vision model to the systems a business already runs, so that a specific task gets done inside an existing process. It is plumbing plus evaluation, and the evaluation is the part that separates a working system from a convincing demonstration.

Next step

Send the task and 50 real examples of it.

You get a measured accuracy figure per field, a cost per document, the review load it implies, and a written recommendation that is allowed to say do not build.

Read by P J Divya, our CTO, not a form queue. Reply usually within one working day, in your time zone.

FROM THE PROTAX LIBRARY

Recent Articles & Guides

Practical reading on business records, tax filings and the decisions that come with running a company.

Newsletter

Stay ahead on compliance, tax & business updates

Get selected business and compliance updates by email. Once a month, no spam.

  • 100% privacy
  • 1 email / month
  • Unsubscribe anytime
Contact ProTax
Chosen by individuals and businesses

Get Help With Your Business

Tell us what you need help with. Our team will review your enquiry and explain the next steps.

Free Consultation No Obligations Expert Advice
FREE Consultation No obligation, no lock-in

Talk To ProTax

Share Your Business Requirement

Instant Response 100% Confidential Expert Advice
FREE Consultation No obligation, no lock-in

Request a Free Quote

Share Your Business Requirement

Instant Response 100% Confidential Expert Advice
AI with a measured accuracy figure, not a demo Test set review threshold