Software consultancy · Toronto & Pune

Paper-based process in. Audited automation out.

We are a consultancy for small and midsize organisations that still run on forms, email attachments and manual re-keying. We map the process, automate the parts that should not need a person, and leave an audit trail on every step a machine takes.

Contact us How we work
Paperdrop — practice data Sheet 01
No. Property Value Remark
01 Combined industry experience 50 years Across product, design, architecture and delivery, carried end to end by a very small team
02 Who we build for SMB & mid-market Owner-led firms, ops teams, internal IT
03 Engagement shape Review, then build Scope agreed in writing before any build begins
04 Delivery locations Toronto · Pune Overlapping hours across North America and IST

Figures describe the practice, not a client outcome. Outcome numbers are quoted per engagement, after the review.

02 · Services

Digital transformation

Paper forms, shared inboxes and spreadsheets replaced with systems of record. We inventory the documents, agree what each one is actually for, and retire the ones nobody reads.

  • Process and document audit
  • Systems selection and integration
  • Migration off paper and PDFs

About digital transformation →

Process automation

The repetitive middle of a workflow — intake, routing, approval chasing, re-keying — handed to software, with exceptions escalated to a named person rather than swallowed.

  • Workflow and approval routing
  • Document intake and extraction
  • Exception handling and alerting

About process automation →

Consulting & training

Advice and hands-on training so the system keeps running after we leave. We write the runbook, train the operators, and hand over the source.

  • Technical due diligence
  • Operator and admin training
  • Handover, runbooks and source

About consulting & training →

03 · AI capability

Automation you can audit

We use language models where they measurably remove manual effort — reading documents, classifying requests, drafting the routine reply — and nowhere else. Every model step is bounded by a lever a human can pull: a confidence threshold, a review queue, a log, a rollback.

That is the same discipline we apply to our own practice. Automating our internal delivery work is why a small team can carry engagements of this size at this price — and it is the capability we hand to clients rather than describe.

  • 01Integration with Copilot, ChatGPT and comparable platforms already licensed in your business
  • 02Domain-tuned assistants grounded in your own documents and terminology
  • 03Context-aware copilots inside the software your team already opens each morning
  • 04Evaluation harnesses, so a model change is a measured change and not a surprise
  • 05Offline AI: open models run on your own hardware, so documents never leave your network
Fig. 01 — a bounded model step
Intake
Extract
Model
Act
Audit levers
Confidence threshold Human review queue Immutable decision log One-step rollback

No step reaches Act unattended unless it clears the threshold; everything below it queues for a person, and every outcome is logged either way.

Models and platforms we have shipped with
  • Claude logoClaudeAnthropic
  • OpenRouter logoOpenRouterMulti-model gateway
  • Ollama logoOllamaLocal & on-premise models
  • Moonshot AI logoMoonshot AIKimi models
  • OpenAI logoOpenAIGPT & ChatGPT
  • Microsoft Azure logoMicrosoft AzureAzure OpenAI & Copilot

Hosted or offline, depending on where your data is allowed to go. Logos are the property of their owners and indicate delivered work, not a partnership.

Beredo, the document-collection product we built, runs on this pattern: the assistant drafts the process and never reads what clients submit.

04 · How we work
01

Review

A short, scoped engagement to start. We sit with the people doing the work and document the process as it really runs.

02

Scope

A written scope with a fixed price, the assumptions it rests on, and what we would not build. You can take it to another firm.

03

Build

Short increments, something usable at the end of each. Automation ships behind its levers, in review mode, before it runs on its own.

04

Hand over

Source, runbook and training to your team. Ongoing support is a separate agreement, not a dependency we design in.

05 · Questions

Frequently asked

The questions we hear in the first conversation. If yours is not here, ask it through the contact form and we will answer it directly.

How long does a process review take?

A review is deliberately short. We sit with the people doing the work, document the process as it actually runs, and return a written scope. How long that takes depends on how many teams and documents are involved, so we agree the length and price of the review itself before it starts.

How is pricing structured?

The review is priced on its own. It ends with a written scope that carries a fixed price, the assumptions it rests on, and a list of what we would not build. You can take that scope to another firm. Ongoing support after handover is a separate agreement if you want one.

What happens when the automation meets something it does not recognise?

It stops and asks. Every model step runs behind a confidence threshold; anything below it queues for a named person with the context attached, and every outcome, human or machine, is written to an immutable log and can be rolled back in one step. Nothing is silently dropped.

Does our data leave our network?

Only if you decide it can. Where documents must stay on your premises or in your country, we run open models on your own hardware and integrate with the local regions of the platforms you already license. Beredo, our own product, is built so the assistant never reads what clients submit.

Which AI platforms do you integrate with?

The ones already licensed in your business where possible: Microsoft Copilot and Azure OpenAI, ChatGPT and the OpenAI API, Anthropic's Claude, and multi-model gateways such as OpenRouter. For offline work we run open models through Ollama. We have shipped with each of these.

Do we have to replace our existing systems?

Usually not. Digital transformation starts with an inventory of every document, form and spreadsheet in the process and a decision about what each one is for. We retire what nobody reads, keep the systems that are sound, and integrate rather than replace wherever we can.

What do we receive at handover?

Source code, configuration, a runbook for day-to-day operation and exception handling, and training for the operators and administrators who will run it. The system is yours to run without us. We would rather leave a trained team than a support dependency.

Who do you work with?

Small and midsize organisations: owner-led firms, operations teams and internal IT whose processes still run on paper, attachments and manual re-keying. We are a very small team by design and take fewer engagements than we could, so the people you meet in the review are the people who build.

Where are you based, and what hours do you work?

Toronto, Canada and Pune, India. The two offices give overlapping coverage across North American and Indian working hours, so a review held in Toronto in the afternoon can be built against in Pune the same evening. See the Toronto and Pune pages.

How do we get started?

Describe the workflow that costs you the most time through the contact form. We reply with what we would look at first and an honest view on whether a review is worth your money.

06 · About

A very small team, on purpose

Paperdrop was formed by product managers, trainers, designers, architects and developers with fifty years of combined industry experience. We work with small and midsize businesses whose processes still run on paper and attachments — work that is slow, error-prone and more expensive than it looks on the ledger.

A conventional delivery team is a chain of verticals: analysis hands to design, design to build, build to test, test to documentation. Every handover costs time and loses detail. We use AI to collapse that chain. The same few people carry an engagement from process review to runbook, with language models doing the drafting, scaffolding, test generation and documentation that used to need a separate role each.

That is why we stay very small and take fewer engagements than we could. The people you meet in the review are the people who build the thing, and there is nobody in between.

Fig. 02 — delivery verticals, collapsed
Conventional team
AnalyseDesignBuildTestDocument

Five roles, four handovers. Each one costs time and loses detail.

Paperdrop
The same few people, review to handover
Drafting specs & runbooks Scaffolding code Generating tests Writing documentation

The grey work is carried by language models and reviewed by the person who owns the engagement. No handovers.

Disciplines, carried by the same few people
Product management Solution architecture Software engineering Interaction & service design Training & change management
07 · Product
Beredo

Set your process up once. Run it forever.

Beredo is the product we built for firms that collect documents from clients. A mortgage application, a tax return, a new-client onboarding: the same steps, the same documents, the same signatures, every time — and every time rebuilt by hand in somebody's inbox. The work is not hard. Keeping track of it is.

You write the process down once: who does what, in what order. Every client after that runs on it. Each person is asked for one thing at a time, in the right order, and the firm stops carrying the process in its head.

  • 01Your clients never make an account. They follow a link from an email and do the one thing waiting on them.
  • 02It chases people so you do not have to, and tells somebody else if it keeps waiting.
  • 03Unlimited people on every plan. You are not charged for your own customers.
  • 04Built for mortgage brokers, accountants, immigration lawyers, property managers, recruiters, clinics and practices.
See Beredo at beredo.io
How it works
01

Build the flow

Steps, and the forms, uploads, signatures and approvals in each. Say who each one belongs to and what counts as done.

02

Start an application

One mortgage, one tax return, one onboarding. Invite the people involved; they need no account and no software, just the link they were emailed.

03

It runs

Each person sees the one thing waiting on them and nothing else. When they finish, the next thing opens for whoever it belongs to, and the firm can see where every application has got to.

Trust, for firms handing over identity documents
Every file and answer encrypted under its own key Deleting destroys the key, not just the row Audit trail of field names, never values Signed PDFs carry a certificate anyone can verify

Priced by how many applications you have open at once, not per person and not per start. Finish one and the space is yours again.

08 · Contact

Describe the process

Tell us about the workflow that costs you the most time, through the contact form. We will reply with what we would look at first, and whether a review is worth your money.

Contact us
Toronto +1 437 449 6106
Locations Toronto, Canada · Pune, India
09 · Careers

Open applications

We hire irregularly and only for work we already have. If you build automation, integrate systems, or teach people to use them — in Toronto or Pune — send a note and something you have shipped. We answer every application.

Send an application

Tell us which discipline you work in, where you are based, and one thing you have taken from paper to production.