// prepared for sinnadurai mahendranathan · may 2026
வணக்கம் ஐயா
// vanakkam ayya · a respectful greeting
A platform that turns 38 years of practice into something that outlasts the building.
You sold MP Accounting. Good — that's the right move at the right time. But the 38 years of context, instinct, and CRA fluency that built it doesn't have to retire with the building. There's a way for it to keep working — for you, on your terms, without you having to lift much of anything.
பிறந்தநாள் வாழ்த்துக்கள்
// a little belated · may 15 was noted
Three reasons this lands on your desk before anyone else's.
01you helped Jason's family. This is Jason giving back, in the form he can.
02the platform serves your kind of practice first — small-to-mid Canadian books, T1 + T2 + HST + payroll, run by RPAs and CPAs in the GTA Tamil professional network you spent four decades inside.
03the train is leaving the station. You have the seat next to him if you want it. If not, it goes to someone on the list at the end of this page.
// section 02 · what the platform does
A library of atoms. A scenario is a composition.
Every accounting deliverable decomposes into small, versioned, self-tested modules. New scenario = new composition, not new code. After Phase N, the marginal cost of one more scenario approaches zero. The atom library is the proprietary asset. Click any family below to see its atoms.
Five real client-to-agent dialogs. Three English. Two Tamil. Each shows the customer voice, the agent response, and the moment your overlay (the accountant view) flags what needs human sign-off. Click any to see the scene-by-scene breakdown.
// audio renders via ElevenLabs · scripts ready · next pass plays in browser
// section 04 · the three surfaces
We don't replace QuickBooks. We compose it.
The platform sits above the existing accounting software ecosystem — not in place of it. Three actors. Three surfaces. One coherent platform.
You already know how QuickBooks works. You already know how Xero works. You already know how TaxCycle works. We are not building another one of those. We are building the layer that connects all three to each other — and to the customer's voice, and to the documents on the customer's phone, and to your sign-off, and to CRA.
// surface 01
Customer
// individual filer · SMB owner
Voice — duplex agentic dialog
Phone photo — every slip, every receipt
SMS — for clients who won't install an app
Bank connect through their chosen platform's OAuth (we never touch credentials)
Clean financial view in their chosen platform
// three pathways
A · Zoho Free entry tier (no software cost)
B · Xero Starter upgrade ($25/mo · best bank feeds + Hubdoc)
C · QBO BYOB (they already have it)
// surface 02
Accountant
// the practitioner overlay
Multi-client dashboard
AI confidence flags — HIGH / MEDIUM / LOW
R-03 sign-off gate — non-negotiable
12-minute review per file vs traditional 90
Full audit trail per atom, per decision
// primary: Xero Practice Manager (free for Silver+ partner)
// fallback for heavy T2: Intuit Accountant Suite (Pro Tax + GIFI Workpapers)
// surface 03
Government
// CRA · IRS · state agencies
T1 / T2 EFILE submission to CRA
1040 / 1120-S / 1065 MeF submission to IRS
T4 / T5 / T4A slip XML
GST/HST + payroll remittance
// CA: TaxCycle (Xero-owned · only CA software with public Developer API)
// US: GoSystem Tax RS (Thomson Reuters · documented REST API)
// Phase-2: own CRA + IRS certification (~$0 govt fees + 7-16 dev-weeks per jurisdiction)
// section 04½ · not starting from zero
This isn't a research project.
The hardest pieces of this platform — the duplex voice agent, the document ingestion, the real-time agent orchestration, the privacy-as-architecture posture — aren't being invented for this. They are already running, in production, in other Obsidian Delta products. What's new here is the application to accounting, not the underlying stack. Click each piece to see what it already does, and what it will do here.
// voice agent
Duplex agentic conversation
Already shipped. Real-time bidirectional voice with contextual back-and-forth — not voicemail, not a chatbot. Used today in Obsidian Delta's existing voice product.
click for what carries over →
// doc ingestion
Phone photo + SMS pipeline
Customer takes a picture, agent reads it in seconds. No app install. Texted documents accepted as natively as uploaded ones. Already proven.
click for what carries over →
// orchestration
Multi-agent atom dispatch
Sub-agents fan out, run in parallel, return into a synthesis pass. The same pattern that drove building this very pitch deck for you.
click for what carries over →
// privacy
Architecture-level data posture
No sell. No share. No train. Encoded in the system design, not in a policy document. Already the standing posture across every Obsidian Delta product.
click for what carries over →
// bridge
Host-to-cloud control plane
Real-time bridge between desktop, cloud services, and external APIs. The same plumbing that lets us ship this site without manual deploy steps.
click for what carries over →
// cognition
Reasoning core + decision gates
The same engine that runs other Obsidian Delta products. R-03 sign-off, confidence flagging, audit-trail logging — all native primitives, not bolt-on features.
click for what carries over →
What's new here is the accounting layer. The atoms in Section 02. The hybrid Zoho-Xero-QBO surface in Section 04. The TaxCycle + GoSystem licensed-engine bridge. The CRA + IRS regulatory framework in Section 09. Everything else is already running. The capital ladder is funding the application — not the underlying engine.
// section 05 · the numbers
What the platform delivers, in dollars.
From the financial analysis pass — customer savings, practice impact, market sizing. Every number sourced from CRA / IRS / Census / Scout 09 fee benchmarks and Scout 13 practitioner economics.
$900M
// canadian customer savings · year
Annual professional-fee savings at 5% Canadian market penetration. CAD.
$6.0B
// us customer savings · year
Annual professional-fee savings at 5% US market penetration. USD.
+$92K
// practitioner net income lift · year 3
For a 300-client RPA practice integrating the platform · semi-active scenario.
+$765K
// succession value uplift
Practice exit multiple shifts from 1.0× revenue (owner-dependent) to 2.5× (platform-served + transferable).
Customer savings · per scenario · annual
scenario
jurisdiction
current cpa fee
platform price
annual saving
saving %
Multi-source individual (T4+T5+RRSP+Sch 3)
CA
$450
$149
$301
67%
T4 + rental T776
CA
$350
$149
$201
57%
Self-employed T2125
CA
$300
$149
$151
50%
CCPC $100K-500K full-service
CA
$9,700
$2,250
$7,450
77%
Sole prop $30K-100K (HST registered)
CA
$5,200
$600
$4,600
88%
Gig worker (W-2 + 1099)
US
$491
$104
$387
79%
S-corp $50K-250K full-service
US
$9,500
$1,600
$7,900
83%
Itemizing homeowner
US
$382
$69
$313
82%
Practitioner impact · 300-client RPA archetype · 3 years
path
year 0
year 3 net
absolute lift
% lift
extra dimension
A · Semi-active hold · keep 300 clients, cut hours
Milestone-based, not pitch-deck-based. One thousand dollars CAD marks the first one. It's not an investment in a business plan. It's a friend saying yes to a friend's project that's already in motion.
// milestone 01 · founding sponsor
$1,000
canadian dollars
Marks the official kickoff of Obsidian Delta's accounting automation Phase 1.
Triggers: Stripe + accounting infrastructure setup, first scout cohort budget,
the engagement letter atom that pulls every subsequent client onboarding.
Status visible in Section 07 phasing.
For being first out of the gate, the standing offer is this:
the Jeyanandan guest room in Bathurst, New Brunswick is yours whenever you want it.
Amma's rasam. Amma's sambar. Amma's appalam.
A front-row seat to the build, watched from the couch — the Atlantic five minutes away if you need a walk.
Stay a weekend. Stay a week. Stay two. Complain about the architecture gently.
Watch how the platform handles a real client file. Comment when it gets something wrong.
That's the whole obligation.
The milestone ladder
Each subsequent milestone is gated by a concrete deliverable. You're not asked to commit to the whole ladder. You're asked to look at the first rung.
// milestone 01
$1K
CAD · current
Platform kickoff, infra setup, first 5 scout cohort completed (already done).
// milestone 02
$10K
CAD
First end-to-end T1 filed through the platform under a real CRA EFILE credential. Demo recorded.
// milestone 03
$25K
CAD
First T2 + GIFI Schedule 100/125 round-trip. First 10 paying customers onboarded.
// milestone 04
$50K
CAD
CCA engine internalized, first practitioner cohort (3-5 firms) onboarded, 100 returns/yr run rate.
None of these ask you to work hard. You've earned the right not to. Pick whichever fits the rhythm of where you are now.
// note · the original four-lane matrix included "A · Practitioner Partner" — your practice anchoring as the platform's first cohort. Since MP Accounting has been sold, that option no longer applies and has been removed. D · Founding Sponsor is what we're hoping for. The other two remain on the table if they fit better.
// section 08 · the phased plan
Five phases. Each milestone-gated.
Phase 1 starts with the $1,000 milestone. Each subsequent phase is gated by a deliverable, not a calendar. Click any phase to see the internalization milestones (what we replace, what we keep licensed) per Scout 17 roadmap.
US launch via practitioner channel. Own EFIN + MeF certification. CRA slip XML internalized.
trigger → CA channel profitable + US partnerships locked
// phase 05
Full Sovereignty
DTC US. Own payroll engines (CA + US). Direct bank integration via Plaid. Zero licensed dependency.
trigger → internalization roadmap complete
// section 09 · the regulatory posture
Privacy is architecture, not policy.
Three load-bearing rules of the platform that match the bar a 38-year practitioner expects.
01R-03 sign-off gate · no return transmits to CRA or IRS without a practitioner's review. The platform does the work. The platform never files without a human signing.
02§7216 client-consent gate · no client data is sent to a licensed engine until consent is captured in the engagement letter atom. Criminal exposure on misuse — handled at the atom layer.
03Full audit trail · every atom run, every voice transcript, every document version, every accountant decision is logged for audit defense.
Liability under IRC §6694, ITA s.163.2, AICPA + CPA Canada conduct codes, PIPEDA, Quebec Law 25 — mapped per scenario in Scout 15 (30,000 words, available on request). The architecture takes the strict-default position on every interpretive ambiguity. We do not sell, share, or train on client data. No exceptions. No enterprise-tier loosenings. The posture is the architecture, not the policy.
// section 10 · we really hope it's you
If not — here's who's next.
No threat in this list. Any one of these would be a fortunate landing spot for the platform. We're showing you the list because we want you to see that the train really is leaving — and because the ones we'd approach next reflect what we think you'd respect about how this gets built. Click any to see why they're on the list.
// section 11 · closer
Whatever you decide, the call is yours.
If yes — coffee at a place you know, when it's convenient, with the Stripe milestone done from the link above. If you also want the rasam offer, amma will make it whenever. If no — congratulations on 38 years, and thank you for everything you've already done for this family. There is no wrong answer here.