Byron
Also known as: usebyron
AI agent platform for business tax preparation that automates the CPA workflow from client document intake to review-ready workpapers and returns, with every number source-linked and every step gated by human approval.
Byron is an AI agent platform for business tax preparation, built for CPA firms preparing 1065, 1120 and 1120-S returns, that runs the engagement from client request list to review ready workpapers while the firm's reviewers keep control.
Its agents sync client data from accounting software, tax systems, email and document management systems and generate the PBC request list, organize incoming documents into a binder, learn how the firm treated each client's items last year across M-1, M-3, depreciation and state apportionment, apply the same treatment to this year's data while surfacing exceptions, and generate the Excel workpapers in the firm's own tabs, templates and conventions, with federal and state K-1s, K-3s and footnotes handled in full.
Excel stays central and synced both ways, so reviewer edits flow back into Byron and into the tax engine without rekeying, and the agents stay on the engagement as late K-1s, revised trial balances and client corrections arrive, updating the workpapers they affect and flagging what changed.
Every prepared value links to its source document and carries a confidence score, reviewers start with the items most likely to need attention, and nothing exports to the firm's tax software until the firm reviews, approves and signs off; every action, edit, approval, override, sync event and export is recorded. Byron is SOC 2 Type II compliant, supports single sign on and role based access so the firm controls who sees which clients and engagements, enforces multi factor authentication, hosts all data in the United States, and never trains models on client data.
It integrates with accounting platforms including Canopy and Truss. Founded in 2025 in San Francisco by a team from Amazon's AGI group and Deloitte, Byron launched publicly in May 2026 with a $6.5 million seed round led by Square Peg and is sold by demo. It is a strong fit for firms automating the preparation behind complex business returns with reviewers in control, and does not offer model choice or self hosted deployment.
Vendor details
Canonical URL
https://www.usebyron.com
Category
Enterprise operations agent
Subcategory
AI tax preparation for CPA firms
Funding status
Independent, based in San Francisco, founded in 2025 by Blaze O'Byrne (CEO) and Wilm Kranz, with a team drawn from Amazon's Artificial General Intelligence group and Deloitte. Byron raised a six and a half million dollar seed round led by Square Peg, with participation from Sorenson Capital, Liquid 2 Ventures, and Correlation Ventures. The company began building in September 2025, brought its first clients live in January 2026, and launched publicly in May 2026, with distribution partnerships through platforms including Canopy and Truss.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Byron works across the tools CPA firms already use, including email, file storage, accounting systems, Excel, and tax software, and it moves approved data into existing tax engines without manual rekeying. Edits made in Excel flow back into Byron automatically. The platform integrates with accounting platforms such as Canopy and Truss, which embed Byron's capabilities into their own workflows, implying an integration surface beyond a single application.
In practice
A firm rebuilds prior year workpapers by hand every busy season. Byron rolls them forward into a current year Excel workbook with this year's data and requests whatever client information is missing.
Business returns are too complex and review heavy to trust to generic AI. Byron's task specific agents source link every number and gate each step, bringing the CPA back in whenever something needs a human.
Client data arrives as messy PDFs and spreadsheets. Byron extracts and reconciles it, prepares review ready output, and syncs approved data into the firm's existing tax software.
Sources & related URLs
Research sources
Agentic Index coverage score
8.5 / 14 capabilities · 61%
| Integrations & Tool Calling | Full |
|---|---|
|
Byron syncs from accounting software, tax systems, email and document management systems, and keeps Excel synced both ways, with reviewer edits flowing back into Byron and into the tax engine. Approved output moves into the firm's tax software under role based export permissions. Canopy and Truss are named as integrated platforms, but the tax engines Byron exports to are not named. SourceByron, usebyron.com/business-tax and usebyron.com/securityread 2026-09-06 |
|
| Workflow Orchestration | Full |
|
Agents run a six stage pipeline across the engagement, pulling client data and generating PBC request lists, organizing the binder, learning last year's return, applying book to tax adjustments, generating the Excel workpapers, and review and export. A specialized agent handles each stage, and Byron says its agents use tools and integrations to run the whole job, from gathering documents to preparing the return for review. SourceByron, usebyron.com/business-taxread 2026-09-06 |
|
| Knowledge Grounding & RAG | Full |
|
Byron ingests each client's prior year workpapers and return, learns how the firm treated specific items across M1, M3, depreciation and state apportionment, and maps current year data onto the same structure with the same tabs, templates and conventions, applying similar treatment as last year and surfacing exceptions. That store of the firm's own prior work per client persists across years and is queried on each engagement, and every prepared value traces back to its source document. SourceByron, usebyron.com/business-taxread 2026-09-06 |
|
| Human Oversight & Guardrails | Full |
|
Nothing exports to the firm's tax software until the firm reviews, approves and signs off. In the approval workflow, agents suggest next steps and reviewers approve the work before it moves forward. Review starts from the exceptions most likely to need attention, the review package carries override and sign off history, and every item has a confidence flag, all inside Byron's own review queue. SourceByron, usebyron.com/business-taxread 2026-09-06 |
|
| Security, Identity & Governance | Full |
|
Byron is SOC 2 Type II compliant, audited annually against the Trust Services Criteria with the report shared under NDA, aligns with IRS Publication 4557 and has annual independent penetration testing. The firm controls access, with support for all common single sign on protocols, role based access that decides who can see which clients and engagements, enforced multi factor authentication and role based export permissions. SourceByron, usebyron.com/securityread 2026-09-06 |
|
| Observability & Auditability | Full |
|
Every prepared value traces back to its source document and supporting context, and an audit trail that follows the workbook records every action, edit, comment, approval, override, workbook update, sync event and export, so every value and action has an audit log. SourceByron, usebyron.com/security and usebyron.com/business-taxread 2026-09-06 |
|
| Memory & State Persistence | Partial |
|
A Chat surface takes instructions and completes tasks, carrying conversation state inside a session. There is no memory layer with a stated scope and lifetime. The engagement state agents pick up where they left off is the application's work product, and the prior year treatments for each client are the firm's knowledge, not agent memory. SourceByron, usebyron.com and usebyron.com/business-taxread 2026-09-06 |
|
| Deployment & Data Residency | Not documented |
|
All data is stored and processed in US data centers on AWS, and no data leaves the country. There is no region a customer selects, no customer cloud, on premises or customer tenant option, and no way to choose. SourceByron, usebyron.com/securityread 2026-09-06 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
The specialized agents for book to tax adjustments, depreciation, apportionment and partnership allocations are stages of one business tax pipeline, and Byron sells one platform through two solution pages, Business Tax and K-1 Processing. There is no catalog, template library or named set of agents the customer chooses among, and Byron does not say whether K-1 Processing can be adopted on its own. SourceByron, usebyron.com and usebyron.com/business-taxread 2026-09-06 |
|
| Triggers & Channel Coverage | Full |
|
Agents are long running by design. They stay on the engagement and pick up where they left off as new documents land, sync from accounting software, tax systems, email and document management systems, process new documents as they arrive and flag what changed, and ask clients for what is missing. A Chat surface also takes instructions. Byron does not say whether the firm can configure schedules or further inbound channels. SourceByron, usebyron.com/business-tax and usebyron.comread 2026-09-06 |
|
| Model Flexibility & Routing | Not documented |
|
No model provider is named. Byron's security FAQ refers to its AI providers in the plural, under terms that prohibit training, and the agents use tools such as Python for calculation. Neither customers nor admins can select a model, and the plural is a passing mention, not routing. SourceByron, usebyron.com/securityread 2026-09-06 |
|
| APIs, SDKs & MCP Extensibility | Partial |
|
Byron integrates directly with the accounting platforms Canopy and Truss, and its launch release describes those platforms bringing Byron into their own workflows, which means an outside platform invokes Byron. There is no public API reference, SDK, MCP server or agent card for the Byron platform, the site has no developer link, and the app sits behind a login. The current pages do not say which way the Canopy and Truss integration runs. SourceByron, usebyron.com and the vendor issued release of 27 May 2026read 2026-09-06 |
|
| Testing, Debugging & Optimization | Not documented |
|
There is no evaluation harness, scored test cases, quality gate on a change to an agent or optimization loop. A confidence score on every prepared item routes reviewers to what needs attention, which checks each output instead of testing a change, and the published 99 and 97 percent accuracy figures are Byron's benchmark of its own product, not something the firm can run. SourceByron, usebyron.com/business-tax and usebyron.comread 2026-09-06 |
|
| Browser & Computer Use | Not documented |
|
Agents reach the firm's systems through connections to file storage, accounting systems, email, Excel and tax software, and work inside Byron's own workspace without controlling a browser, desktop or computer. Byron does not say how approved output is pushed into the tax engine. SourceByron, usebyron.com/business-taxread 2026-09-06 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Contact for pricing
What is public
No list prices or entry point are published; Byron is demo gated.
Billing mechanics
Not published. Inference, not stated by the vendor: pricing is likely to scale with firm size or return volume.
Cost watchouts
Busy season volume can raise effective cost if pricing scales with returns processed.
Variable cost rationale
Scope likely scales with the number of returns, entities, or seats a firm processes, so cost can grow with volume, though pricing is negotiated under annual agreements rather than metered by usage.
Additional watchouts
Confirm how pricing scales with returns, entities, and seats, and how access is provided when reached through partner platforms like Canopy or Truss.
Sales call required
Yes, required for paid access
Free / trial
Demo led evaluation; no public free tier
Lowest paid plan
Not published
Key ambiguities
No public anchor for entry price or scope units.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Abacum — FP&A platform for growing companies whose AI Modeler builds, edits…
- Adopt AI — AI native accounting and bookkeeping firm whose agents run…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
Alternatives to Byron
The closest documented capability profiles to Byron among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Basis7.5 / 14A lighter documented profile than ByronByron vs Basis →
- Otel AI8.5 / 14Adds documented Model Flexibility & Routing
- RegASK8.5 / 14Adds documented Deployment & Data Residency
- Tabs8.5 / 14Fuller documented coverage on APIs, SDKs & MCP Extensibility
- Vic.ai7.5 / 14A lighter documented profile than Byron
- Warp9.5 / 14Fuller documented coverage on Prebuilt Agents, Templates & Packs and APIs, SDKs & MCP Extensibility
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded