T1 Client Intake for Accounting Firms: The Process, the Software, and What AI Agents Change

Ask a Canadian firm where tax season actually hurts and the answer is never the tax calculation. It's intake: getting each client's documents in, understanding what's there, noticing what's missing, chasing it down, and knowing - with confidence, across hundreds of files at once - which returns are ready to prepare and which are still waiting on a slip. Calculations are instantaneous; intake is weeks, and it's the stage where files stall, staff burn out, and errors are born.

This guide lays out what a good T1 intake process looks like, what to demand from client intake software, and how AI agents are changing the economics of the whole stage.

What a T1 intake process actually has to do

Strip away the tools and every firm's intake runs the same eight steps: open the engagement, request documents from the client, collect what arrives, identify each document, record its data, detect what's missing or inconsistent, follow up until the gaps close, and hand a complete package to preparation. Simple to list, brutal at volume - because steps three through seven don't happen once per client. Real intake is a long-running conversation. The client sends four documents in February, two more in March, and answers a question the day before the deadline. Multiply by five hundred clients and the real problem isn't any single step; it's state - knowing exactly where every file stands, every day, without anyone reconstructing it from an inbox.

Why the traditional toolkit breaks

Most firms run intake on email, a shared folder, a spreadsheet tracker, and a static client questionnaire. Each piece fails in a characteristic way. Email buries documents in threads and makes the follow-up trail personal to whoever sent it. Folders hold files but understand nothing about them - a folder can't tell you the T5 is missing. Spreadsheet trackers rot within weeks because they're updated by hand, by busy people. And the annual questionnaire under-collects by design: a static form asks everyone everything, gets skimmed, and still can't ask the one specific question that matters ("your donation list shows CityKidz with no amount - what was it?").

Client portals improved the collection step - documents arrive in one place, with reminders. But a portal is a mailbox with a checklist. It collects; it doesn't understand. Everything after upload - reading the slip, recording the data, spotting the gap, chasing it - is still people.

What to look for in T1 client intake software

Whether or not agents are on your roadmap, the evaluation checklist has changed. Six capabilities separate intake software from document storage:

A structured tax data model, not a file cabinet. The system should know what a T4 is - its boxes, its rules, its relationship to the return - so a document becomes recorded, validated data, not a PDF waiting for re-keying.

Validation at the point of entry. Required fields, cross-field rules, and slip-level checks should run when data is recorded, so problems surface in February, not at review in April.

Gap tracking as a first-class feature. "What's missing" should be a computed fact with an open item attached to the exact record - not a feeling, and not a column in somebody's spreadsheet.

Provenance on every number. Each recorded figure should link to the source document it came from, with the document stored alongside. This is what makes the file reviewable later without re-checking every line.

A completeness gate. The system - not the preparer's memory - should decide when intake is done: every applicable section addressed, every required field present, every open item closed, and a workflow status that won't advance until that's true.

Clean handoff. Intake feeds preparation. The output should be a structured, validated package your tax prep software and practice management stack can consume, not another folder.

The questionnaire, rethought

The static T1 questionnaire deserves its own funeral. Its replacement isn't a better form; it's gap-driven asking. When intake runs on a structured data model, the system knows precisely what's missing per client - so the client gets one short, specific request ("Box 24 and 26 from your ABC T4; the CityKidz donation amount and receipt") instead of a forty-question form. Specific asks get faster, better answers, and they stop training clients to ignore your emails.

What AI agents change

Intake is the most automatable stage of tax work, because most of it is reading, recording, comparing, and following up - exactly what modern AI agents do well. An agent can watch the portal, read each document as it arrives, record the data into the intake system, open the gap items, and send the specific follow-up request, running the whole loop daily across every client without fatigue.

But - and this is the part firms evaluating "AI intake" tools should press hardest on - the agent is the easy half. An agent with no governed system underneath will save extracted numbers into a spreadsheet, invent a value when a required field blocks it, and lose track of state between sessions. Everything in the checklist above is what makes agent-run intake safe: the data model gives the agent something structured to write into, validation catches its errors, provenance makes its work reviewable, gap tracking gives it honest ways to record what it doesn't know, and the completeness gate stops it - like any staff member - from calling a file done early.

Where Armada T1 fits

Blackspark builds Armada T1, agent-ready infrastructure for Canadian T1 workflows - the intake layer described above, exposed through APIs and MCP so AI agents can operate it under governance. The Canadian tax data model, field-level validation, document provenance, task tracking, and completeness gates are the platform; your agents (or your staff) do the work on top. Files hand off cleanly to the tax preparation and practice management software your firm already runs.

If you're rethinking intake for next season - with agents or ahead of them - request access on our homepage.

The bottom line

T1 intake is a state-management problem wearing a document-collection costume. Firms that treat it as filing - folders, emails, trackers - pay for it every March and April in overtime and missed slips. Firms that treat it as structured data with computed completeness get something new: an intake stage that can largely run itself, with people handling only the judgment calls. That's the standard to evaluate any intake solution against this year, because it's the standard AI agents are about to make normal.

Run Your Practice on Karbon? Your T1 Intake Can Now Run Itself.

If your firm runs on Karbon, your practice management is already in good shape: work items, templates, client tasks, a portal your clients actually use. What Karbon was never built to do is the tax work itself — reading the donation list that just landed in the portal, getting it into a tax file, noticing that one amount is blank, asking the client for exactly that, and flagging the thing that needs a preparer's judgment to a preparer instead.

That gap is where your team's tax season goes. And it's now work an AI agent can do — inside the Karbon workflow you already have.

We know because we tested it, end to end, against a live Karbon tenant. This post describes what that looks like for a firm, in plain terms.

One instruction

A preparer opens Claude — the AI assistant from Anthropic that's been getting so much attention across professional services — and types:

"Start preparing a 2025 T1 for Jill Harvey."

Here's what happens next, with no one touching a keyboard:

The agent finds Jill in Karbon and checks for documents she's already uploaded. It creates the T1 work item from your firm's own template — the same checklist your staff would use. The template fires the standard document request to Jill through the Karbon client portal, so she gets the familiar email and task, exactly as if your admin had sent it. When Jill's documents arrive, the agent picks them up from the portal, reads them, and records every figure into a structured tax file in Armada T1 — each one linked to the exact document it came from.

Then the part that separates this from a demo. Among Jill's uploads is the document every preparer knows: a homemade spreadsheet of her donations. Four charities. One amount blank. And a typo — "Canadia Red Cross."

Watch what the agent does with that, because this is the whole product in one moment:

It records all four donations exactly as written, each linked to the stored spreadsheet. For the missing CityKidz amount, it doesn't guess — that's information only Jill has. So it sends a follow-up request through Karbon, and Jill gets a portal task and an email asking for precisely that: "Your 2025 T1 — CityKidz donation: amount and receipt needed." For "Canadia Red Cross," it doesn't silently correct the spelling, and it doesn't pester Jill about it either — whether that's the Canadian Red Cross is a preparer's call, so a note lands in your Karbon triage instead.

Two gaps. Two different destinations. Both the right ones. And when Jill answers her portal task, the reply lands exactly where the agent is already watching.

Every step of that flow ran against live systems. Not a concept video. An AI agent working in between two separate systems, without any hard-coded integration.

The question you should be asking

Couldn't you do this with just Claude and Karbon? Claude reads documents beautifully, and Karbon has an API.

You could build that demo in a weekend. Here's what it can't survive: Karbon is a system of record for your practice — clients, work, communication. It is not a system of record for a tax return. When Claude reads Jill's donation list, the figures have to go somewhere. Without a tax platform in the loop, "somewhere" is a spreadsheet, a document, or the agent's own memory of the conversation. That fails in ways that matter to a professional practice:

Nothing checks the work. A donation record needs an organization and an amount. A 2025 T4 has more than eighty fields with CRA box mappings and rules connecting them. A spreadsheet knows none of that. Armada T1 does — every entry the agent records is validated against the actual Canadian tax data model, for the actual tax year, and the blank CityKidz amount is tracked as an explicit, open item from the moment it's recorded.

Nothing stops invention. This is the risk nobody sees until it bites. An AI that can't save incomplete work will eventually fill a blank with something plausible — a $75 that was never on any receipt. Armada is designed for the opposite: an honest partial record always saves, every gap is tracked field by field, and the agent never has a reason to make a number up. A blank on the client's list stays blank in the file — with the follow-up already on its way to the client.

Nothing proves where numbers came from. When your reviewer — or the CRA — asks about a donation claim, "the AI put it in a spreadsheet" is not an answer. In Armada, every record links to its source document, and the document itself is stored with the file. The reviewer sees the four donations, clicks through to Jill's actual list, done.

Nothing says when it's finished. In a spreadsheet, "intake is complete" is a feeling. Armada computes it — and its workflow gate refuses to advance a file with open gaps, whether an agent or a human is pushing. We watched the agent try. The platform said no. That refusal is the product.

That's the division of labour: Claude does the work, Karbon runs the practice, Armada T1 holds the return — and governs it.

What setup actually involves

No developers, no integration project, no middleware to buy:

• Karbon: generate an API key from Settings (a five-minute, built-in feature), and clone your T1 work template once so the document request sends automatically when a work item is created. Your existing template, your existing client experience.

• Claude: your firm's Claude plan, with the Armada T1 MCP server added — a configuration step, not a development one.

• Armada T1: an account for your firm.

There is no code to maintain, because there is no hard-coded integration. The agent is the integration. When we changed our minds mid-test about where documents should live, nothing had to be rebuilt — the agent adapted, the same afternoon.

What stays human

Your people keep every judgment call — starting with whether "Canadia Red Cross" is a typo or a charity you've never heard of. They review the file, advise the client, sign off, file. What they stop doing is downloading, re-keying, cross-checking, and writing "just following up on your donation receipts" emails. The agent handles the chase; the platform guarantees nothing advances until the inputs are clean; your preparers work the exceptions that surface in Karbon triage — the same place their work already lives.

For a Karbon firm, this isn't a new system to learn. Your clients see the same portal. Your staff see the same work items. The difference is who — or what — is doing the grunt work in between.

If you run T1s on Karbon and want to be among the first firms to put an agent on intake, request access on our homepage.

This post provides general information for tax professionals and is not tax, legal, or filing advice. Integration behaviour verified against live Karbon and Armada T1 environments, August 2026; Karbon is a trademark of Karbon, Inc., and no partnership or endorsement is implied.

Starting a Tax Preparation Business in the Agentic AI Era

There has never been a cheaper time to start a financial services practice, and there has never been a better founding service than tax preparation. Those two claims are related, and this post is the argument for both, aimed at the person we keep meeting: early in their career, credentialed or working on it, ambitious, and wondering whether to join an established firm or build something of their own.

A generation that wants to build

If that's you, you have company. An RBC poll found that 59 percent of Canadians aspire to own a business, the highest level since 2017, with more than 2.6 million already working for themselves. BDC's February 2026 study of 1,505 self-employed Canadians found the youngest cohort is the most growth-hungry of the group. And the intent is converting into action: Futurpreneur, which finances entrepreneurs aged 18 to 39, saw applications jump 50 percent year over year nationwide last fiscal quarter, and 65 percent in Alberta. Its CEO credits changing expectations around traditional employment.

There's a caution inside the same data. Statistics Canada found that Canadians aged 15 to 34 post the highest entry rates into incorporated business ownership, and also the highest exit rates. Plenty of young founders start; fewer build something that lasts. The difference is rarely talent. It's usually the business model: whether revenue recurs, whether demand has to be manufactured, and whether the founder can serve enough clients before the runway ends.

Which is exactly why the founding service you pick matters more than almost any other decision.

Why tax preparation is the keystone service

Start with the business logic, because it hasn't changed in fifty years and AI doesn't change it either.

A tax return is the one financial service almost every adult needs every single year, on a deadline the government enforces for you. Nobody needs to be convinced to file; they only need to be convinced to file with you. That is a far easier sale than any other financial service, where the first job is persuading someone they need the product at all. You still have to win clients, but you're competing for demand that already exists and renews annually by law.

More importantly, look at what you hold after preparing someone's return. Their income and how it's earned. Their RRSP room and whether they use it. Their dependants. Their donations. Their side business and what it actually makes. Their capital gains, their debts surfacing through interest deductions, their marital changes, their approaching retirement. A completed T1 is the most complete financial snapshot of a household that any professional ever legitimately assembles.

Every other financial service is downstream of that snapshot. Retirement planning starts with the RRSP room you can see. Insurance conversations start with the dependants and the mortgage interest you can see. Investment advice starts with the unregistered account activity you can see. Firms have always known this: tax preparation is rarely the most profitable service in the building, but it is the service that makes every profitable conversation possible.

The catch, historically, was that the keystone came with a grind attached. Tax season meant gathering documents, keying slips, chasing clients for the forms they forgot, and re-checking everything under deadline pressure. To hold five hundred client relationships you needed a staffed back office, and to afford a staffed back office you needed five hundred clients. That loop is what kept new entrants out.

The grind is now delegable

That loop just broke, and this is the part that matters if you're starting today.

AI agents can now do the intake grind: read the documents clients send, build structured tax records, notice what's missing, chase it politely and persistently, and assemble reviewer-ready files. Not as a demo, but as governed workflow, provided the agents work on infrastructure built for them, with validation on every write, provenance on every number, and tracked gaps instead of invented ones.

That's what we build. Armada T1 is agent-ready infrastructure for Canadian T1 workflows: AI agents operate the intake, and the platform governs what they're allowed to do and what counts as done. A morning agent sweeps the intake folder and organizes new client documents. A chase agent works the missing-items list. A review agent packages completed files, and a handoff agent moves them into filing software. Those aren't hypothetical job titles. They're the workflow roles a small firm staffs with people, available to you as software.

The strategic consequence is simple: the back-office loop that kept solo practitioners small no longer binds. One licensed professional with an agent-first setup can hold a client book that used to require a team, because the work that required the team is exactly the work agents do best.

Agent-first beats agent-later

Here's the advantage you have over every established firm, and it's worth taking seriously because it's temporary.

Existing practices are retrofitting. They have desktop software, staff trained on manual process, folders full of habits, and busy seasons that punish experimentation. Most will automate slowly, around the edges, without changing shape. You have no legacy process to protect. You can design the practice around the agents from day one: intake that lands in structured, validated records instead of a shared drive; follow-ups that happen automatically instead of when someone remembers; a completeness gate that decides when a file is ready instead of a feeling. You inherit none of the eleven-p.m. re-keying because you never build the workflow that requires it.

Being small stops being a disadvantage. It becomes the moat: you can be agent-first while the incumbents are agent-eventually.

What the agents don't do

An honest list, because the pitch falls apart without it.

Agents don't hold the relationship. The client trusts a person with their finances, and that person is you. Agents don't exercise professional judgment on ambiguous situations, and the good ones are built to escalate rather than guess: the missing box, the receipt that could be two things, the residency question. Those land in your queue, which is the point. Your day is judgment and conversations; the agent's day is everything else.

And agents don't carry your obligations. You still register with the CRA for EFILE, you still meet whatever licensing applies to the services you offer, and if you plan to sell insurance or investment products you need the corresponding provincial licensing before a single conversation happens. AI compresses the grind, not the responsibility. Frankly, that's also why this opportunity is real: the credentials and accountability are the part that can't be automated away, and you're the one bringing them.

The flywheel, from the first client

Put the pieces together and the model looks like this. Tax preparation acquires clients at a lower cost than any other financial service, because the demand already exists and comes back every spring. Agents do the intake, the chasing, and the assembly, so each client costs you minutes of judgment instead of hours of clerical work. The completed return gives you, with the client's consent, the most complete picture of their financial life available anywhere. And that picture is a standing list of conversations worth having: the unused RRSP room, the new baby and no life insurance, the side business ready to incorporate, the retirement five years closer than the portfolio assumes.

Each of those conversations is a planning engagement, an insurance policy, or an investment relationship. Higher margin, deeper trust, and all of it anchored by the keystone service that brings the client back every spring. The old constraint was that you couldn't afford to offer the keystone until you were big. Now the keystone is the cheapest thing you offer.

Starting stack

What this looks like practically: your credentials and registrations, a client-facing identity, filing software for the returns themselves, and Armada T1 as the system of record your agents operate. From there, add clients, not headcount.

We're building for exactly this founder. If you're starting an agent-first practice, request access on our homepage.

This post provides general information and is not tax, legal, licensing, or business advice. Confirm CRA registration requirements and provincial licensing rules for any regulated products before offering services.

Why Not Just Point an Agent at the Intake Folder?

If you build agents for a living, you've already had this thought: tax intake looks easy. Claude or GPT can read a T4 PDF flawlessly. Give an agent access to the client's document folder, a decent prompt, and a spreadsheet to write into, and you've automated intake. You can build that demo in a weekend, and it will genuinely work — in the demo.

We know because that demo is roughly where every conversation with an implementer starts. This post is the list of questions that separate that weekend demo from something a Canadian accounting firm can actually run a tax season on. It's also, not coincidentally, the list of reasons Armada T1 exists.

Where does the data go?

An agent that extracts slip data has to put it somewhere, and the somewhere is the whole game. A spreadsheet or a JSON file has no schema, so every extraction decision the model makes — field names, formats, what counts as box 26 — is improvised per document. Across five hundred returns and three agent versions, you get five hundred slightly different improvisations.

A 2025 T4 has more than eighty typed fields, with CRA box mappings, picklists, and conditional relationships. And it's different from the 2024 T4, because the schema changes every year. Somebody has to own that structure. If it isn't your platform, it's your prompt — which means it's nobody.

What stops the agent from making things up?

This is the question that should keep implementers up at night, and the answer can't be "we prompted it not to."

Here's the mechanism nobody designs for on purpose. Real client documents are incomplete — the T4 arrives without box 26, the receipt is missing a date. If your pipeline validates strictly, records that fail validation don't save. An agent that can't save its work will, sooner or later, fill the gap with something plausible to get past the gate. Strict validation, applied naively to agents, teaches them to fabricate. And a fabricated number in a tax return isn't a UX bug; it's a professional liability event.

The alternative isn't no validation — it's validation designed for how agents actually work. In Armada T1, an honest partial record always saves, and every gap and rule violation is tracked against it as explicit, field-level debt. The agent never faces a choice between losing its work and inventing a number. The gaps then surface at a verify boundary, and an incomplete file cannot advance in the workflow. Truth is enforced by architecture, not by prompt.

Where did this number come from?

A reviewer looking at agent-built work has one overriding question: can I trust it? A spreadsheet row can't answer that. In Armada T1, every record carries a provenance link to the source document it came from — the specific slip, receipt, or CRA import. That single design decision is most of the difference between "the AI filled in a spreadsheet" and "here's an audit-ready file a reviewer can sign off on."

What happens over six weeks?

The demo processes a folder in one sitting. Real intake is a long-running conversation: the client sends four documents in February, two more in March, and answers a question the day before the deadline. A generic agent has a context window; it doesn't have durable state. Armada T1 is the state: tax files with explicit workflow statuses, open tasks anchored to the exact record they're about, and a completeness picture that any agent — or human — can pick up cold, weeks later, mid-stream. The third session doesn't need to re-derive what the first two did.

Who is allowed to do what?

Point an agent at a folder and you've granted it everything, invisibly. There's no identity, no scoping, no record of what it read or wrote. That's fine for a demo and untenable for a firm handling SINs and income data under professional obligations. Armada T1 treats agents as first-class principals: identified as agents, scoped to one firm's tenant, permission-bounded, with every call logged. When the reviewer, the partner, or eventually the regulator asks what the agent did — there's an answer.

Who says it's done?

In the folder-and-spreadsheet version, "intake is complete" is a feeling. In Armada T1 it's a computed fact: a file-level completeness check across every applicable section — missing required fields, rule violations, unresolved sections — and a workflow gate that refuses to advance the file until the inputs are actually clean. The gate applies to agents and humans alike. This is what we mean by "agents operate, Armada governs": the agent does the work; the system decides what counts as finished.

And then what?

Intake isn't the destination. The structured file has to land in the tax preparation and practice management software the firm already runs. A spreadsheet gets re-keyed by a human, which quietly deletes most of the value the agent created. Armada T1's endpoint is a structured, validated, provenance-linked file built for handoff.

The build-vs-buy math

None of the above is exotic. A strong team could build all of it: the versioned Canadian tax data model, the CRA-parity validation rules, the provenance layer, the task system, the workflow gates, the permissions and audit. That's the point — it's not a weekend demo, it's a multi-year system of record in a domain where the schema shifts annually and the cost of a wrong number is measured in reassessments and errors-and-omissions claims.

Here's the thing about the agent layer: harnesses are becoming a commodity. There are hundreds of them, the frontier labs give them away, and every quarter the models get better at operating whatever surface you hand them. The scarce asset in an agentic tax stack isn't the agent — it's the governed surface the agent operates. Our own testing bears this out: a general-purpose model with a two-page instruction file runs a clean, auditable intake on Armada T1, because the guarantees live in the platform, not the prompt.

So build the part where you differentiate: your orchestration, your client experience, your firm-specific workflows. Point it at a system of record that already knows what a T4 is, refuses to let anyone — agent or human — call an incomplete file done, and never gives your agent a reason to invent a number.

That's the division of labour Armada T1 proposes. If you're building on it — as a consultancy, an MSP, or a firm's internal team — request access on our homepage.

This post provides general information for tax professionals and agent implementers and is not tax, legal, or filing advice.

Canadian Tax Preparation Software: What Firms Should Look For in the Agentic AI Era

Canadian tax preparation software has been remarkably stable for two decades. A firm picked a T1 suite — ProFile, TaxCycle, DT Max, Cantax — installed it on desktops, and ran every return through the same keyboard-driven workflow: collect documents by email, re-key slip data, chase missing information by phone, review on screen, EFILE, repeat a thousand times before April 30.

The calculation engines in that software are excellent. The workflow around them is where firms actually lose the hours — and that workflow is what's now changing, because AI agents have become capable of doing large parts of it. This guide looks at the Canadian professional tax software landscape, what traditional suites do and don't cover, and what firms should evaluate as agent-driven automation enters the market.

The Canadian professional tax software landscape

Most Canadian firms prepare T1s in one of a handful of established desktop suites: Intuit ProFile, TaxCycle, Thomson Reuters DT Max, and Wolters Kluwer's Cantax and Taxprep. All are CRA-certified for EFILE, all have mature forms coverage, and all are fundamentally calculation and filing tools: you put clean data in, they compute the return and transmit it.

Consumer products — TurboTax, Wealthsimple Tax, H&R Block's software — serve the do-it-yourself market and aren't built for firm workflows at all: no multi-preparer roles, no client pipeline, no practice-level view of two thousand returns moving through statuses.

The interesting gap sits in front of the professional suites: everything that happens before clean data reaches the calculation engine, and everything that happens around the return afterward.

What traditional tax prep software does well — and where it stops

Traditional Canadian tax preparation software is very good at computing tax and filing returns. What it cannot do is operate the workflow: onboard a client, collect and interpret their documents, notice that a T4 arrived but the RRSP slip didn't, create and chase follow-up tasks, capture where each number came from, assemble a reviewer-ready summary, or move a file through defined stages with an audit trail.

Firms fill that gap with people — plus email chains, shared folders, spreadsheets, and re-keying. During compression season, that manual glue is the bottleneck: preparers spend more time gathering and transcribing than exercising judgment.

What's changing: AI agents in tax workflows

The new generation of AI agents can read documents, extract slip data, ask clients for what's missing, and carry out multi-step work with minimal supervision. But agents can't do that safely inside software designed for a human with a mouse. Screen-driven desktop software gives an agent nothing to hold on to — and no guardrails if it holds on wrong.

What agents need is different: a structured backend with scoped permissions, well-defined programmatic operations, field-level validation, and reliable handoff protocols. In practice that means software exposing its tax data model through APIs and MCP (Model Context Protocol) tools — so an agent can discover what a tax file requires, write records the system validates, flag gaps as tasks, and leave a complete audit trail. The software governs; the agent operates.

What should firms look for in tax filing software for accountants?

Whether or not a firm plans to deploy agents this year, the evaluation criteria have shifted. Six questions worth asking of any Canadian tax preparation software today:

Is it cloud-based? Desktop installs mean data locked to machines, manual updates, and no way to expose safe programmatic access. Cloud architecture is the precondition for everything else on this list.

Does it cover the workflow, or only the return? Calculation and EFILE are table stakes. The expensive hours live in intake, document collection, missing-information follow-up, review preparation, and status tracking.

Is it agent-ready? Ask specifically: are there APIs or MCP tools an AI agent can use? Can permissions be scoped so an agent can gather data but not file? Software that only a human can operate will only ever be as fast as your humans.

Does it validate at the field level? Agent-entered (and human-entered) data should be checked against the tax rules on write — required fields, bounds, cross-field rules — not discovered at review.

Does it keep provenance? Every figure on a return should link back to its source document. That's what makes an agent-built file reviewable and defensible.

Does it hand off cleanly? No firm replaces its entire stack at once. New workflow infrastructure has to co-exist with the EFILE software and practice management tools a firm already runs — structured handoff out, not lock-in.

Where Armada fits

Blackspark builds Armada, agent-ready infrastructure for Canadian T1 workflows — the layer in front of, and around, the calculation engine. Armada exposes clients, tax files, sections, required fields, validation rules, tasks, review states, and filing handoff capabilities through APIs and MCP tools, so AI agents can gather taxpayer information, create structured and validated tax records with document-level provenance, open follow-up tasks for missing information, prepare reviewer-ready summaries, and hand off to the tax prep and practice management software a firm already uses.

The governing principle: agentic workflows should not mean uncontrolled workflows. Agents operate inside scoped permissions; Armada enforces the validation rules, completeness gates, and audit trail. Human reviewers stop doing manual data work and assume oversight roles.

Armada is built for accounting firms, consultants, MSPs, and automation teams implementing AI in tax practices. If that's the problem you're working on, request access on our homepage.

The bottom line

The Canadian tax preparation software market spent twenty years competing on forms coverage and calculation accuracy — a race the incumbents effectively finished. The next decade will be decided by workflow: which platforms let firms deploy AI agents safely against the gathering, validation, and follow-up work that consumes tax season. When you evaluate software now, evaluate it as the system your future agents will operate in — because the firms that get that layer right will prepare more returns with less compression-season pain, at higher and more consistent quality.

This post provides general information for tax professionals and is not tax, legal, or purchasing advice. Evaluate any software against your firm's own requirements, security policies, and CRA obligations.

TOSI Rules: Everything You Need To Know

The tax on split income (TOSI) rules are one of the most consequential — and most frequently misapplied — parts of preparing T1 returns for owners of private corporations and their family members. Since 2018, they have determined whether dividends, trust allocations, and certain other amounts paid to a family member are taxed at that person's own marginal rate or at the top marginal rate, with most personal credits stripped away.

This guide covers how the rules work, who they apply to, the exceptions that matter in practice, and how tax preparation teams can screen for TOSI exposure systematically rather than catching it (or missing it) file by file.

What is income sprinkling?

‍Income sprinkling — often called income splitting — is the practice of directing income from a high-income individual to family members in lower tax brackets, most commonly by paying dividends from a private corporation to a spouse or adult children who hold shares. Before 2018, this could produce substantial family-level tax savings even when the recipients had no involvement in the business.

‍The TOSI rules exist to shut down that version of the strategy. Income splitting itself is not prohibited — several forms of it remain fully legitimate — but income sprinkled to family members who neither work in the business nor put capital at risk is now generally taxed as if the high-income earner had kept it.‍ ‍

What are the TOSI rules?

‍TOSI is found in section 120.4 of the Income Tax Act and has applied in its current form since January 1, 2018, when the former "kiddie tax" on minors was extended to adult family members. The mechanics are blunt: when an amount is "split income" received by a "specified individual" and no exception applies, it is taxed at the top federal marginal rate (33%, plus the top provincial rate), and the recipient loses the benefit of most personal credits against that income.

‍Split income includes, among other things: taxable dividends from private corporations, shareholder benefits, income allocated from a partnership or trust that is derived from a related business, income from certain debt obligations, and certain capital gains on dispositions to non-arm's-length parties.

‍Notably, salary is not split income — reasonable wages paid to a family member for actual work are tested under the ordinary reasonableness rules, not TOSI.‍ ‍

Who do the TOSI rules apply to?

‍TOSI applies to a "specified individual" — essentially any Canadian-resident individual, adult or minor, who receives split income where a related person (the "source individual") is actively involved in the underlying business. The classic fact pattern: one spouse runs an incorporated business, the other spouse and adult children hold dividend-paying shares.

‍Age matters enormously under the rules. Minors are caught in almost every scenario. Recipients aged 18 to 24 face the strictest adult tests. Recipients 25 and over have access to the widest set of exceptions, and recipients whose spouse is 65 or older get a further exception that mirrors pension income splitting.

What is the tax rate under TOSI?

‍Split income caught by TOSI is taxed at the top combined federal-provincial marginal rate regardless of the recipient's other income — federally 33%, with combined rates exceeding 50% in most provinces. The amount is reported on Form T1206 (Tax on Split Income) and flows to line 40424 of the T1. Because the recipient also loses most credits against that income, TOSI routinely produces a worse result than if the source individual had simply earned the income directly.

What are the exceptions to the TOSI rules? (Excluded amounts)

‍The exceptions — "excluded amounts" in the legislation — are where nearly all of the practical analysis happens. The four that matter most:

‍Excluded business (the 20-hour rule). Amounts from a business in which the recipient (18 or older) is actively engaged on a regular, continuous and substantial basis — in the current year or in any five prior years, which need not be consecutive. Working an average of at least 20 hours per week during the part of the year the business operates is deemed to meet the test. This is the workhorse exception for genuinely active family members, and the five-prior-years branch means a retired founder's spouse who put in the hours a decade ago can still qualify.

‍Excluded shares. For recipients 25 or older who personally own shares representing at least 10% of both the votes and the value of the corporation — provided the corporation earns less than 90% of its business income from the provision of services, is not a professional corporation, and derives substantially all its income from its own business rather than a related business. Service businesses and professional corporations are deliberately shut out of this exception, which is why it fails more often than owners expect.

‍Reasonable return. For recipients 25 or older, amounts that represent a reasonable return on the recipient's contributions of labour, capital, and risk assumed, considering historical payments as well. For recipients 18 to 24, only a "safe harbour capital return" (a prescribed-rate return on arm's-length capital they contributed) or a reasonable return on such capital qualifies. Documentation is decisive here — the CRA assesses reasonableness on the facts.

‍Age 65 exception. If the source individual is 65 or older, amounts paid to their spouse or common-law partner are excluded — deliberately aligned with pension income splitting so that business owners are not worse off than pensioners in retirement.

‍Beyond these four, excluded amounts also cover several specific situations: property inherited from a parent (or from anyone, if the recipient is a student or eligible for the disability tax credit), property received on marriage breakdown, taxable capital gains arising on death, and gains on property eligible for the lifetime capital gains exemption (qualified small business corporation shares and qualified farm or fishing property) — the last of which applies even to minors.

How is TOSI different from legitimate income splitting?

‍Several income-splitting strategies remain fully available because they never produce "split income" as defined: pension income splitting between spouses, spousal RRSP contributions, properly structured prescribed-rate loans for investment income, paying family members reasonable salaries for real work, and TFSA contributions funded by gifts between family members. A useful mental model: TOSI polices private-corporation income flowing to relatives; it does not police the ordinary tools Parliament built for household tax planning.

How should tax preparation teams screen for TOSI?

In a firm setting, TOSI errors rarely come from misreading the law — they come from intake gaps. The facts that decide TOSI treatment (hours worked per week, share percentages by votes and value, the corporation's services revenue mix, the recipient's age, the source individual's age) live with the client, not on the slips. A T5 from a private corporation looks identical whether or not TOSI applies.

‍That makes TOSI a workflow problem as much as a technical one. Files with private-corporation dividends should be flagged at intake, the exception-relevant facts gathered as structured data with the client's answers documented, and a T1206 prepared whenever no exception clearly applies. This is exactly the class of problem Blackspark builds for: Armada, our agent-ready infrastructure for Canadian T1 workflows, lets AI agents gather taxpayer information, record it as structured, validated tax data with full provenance, and open follow-up tasks automatically when a required fact — like the hours a family member actually worked — is missing from a file. Firms interested in automating intake-stage screening can request access on our homepage.

The bottom line

TOSI is settled law, not a proposal — it has applied since 2018, it taxes caught income at the top marginal rate, and the difference between a caught dividend and an excluded one usually comes down to facts a preparer must actively collect: hours worked, share structure, revenue mix, and age. Treat those facts as required intake data rather than review-stage discoveries and the rules become manageable; treat them casually and they become reassessments.

This post provides general information for tax professionals and is not tax, legal, or filing advice. Confirm positions against the Income Tax Act, current CRA guidance, and the facts of each client file before applying them.