Xero JAX, Sage Copilot, or a partner build: when each wins for an Irish practice

Three months after Xero opened the Just Ask Xero (JAX) beta in the UK and Sage shipped Copilot in Sage 50 v31.1, every Irish practice partner is fielding the same question from a tech-curious staff member or junior partner: "why aren't we just using what they already ship?"

It is the right question to ask. The honest May 2026 answer is "for a single-system, single-currency, single-entity practice on Xero or Sage 50, you might already be." JAX in beta does meaningful work on invoice drafting, payment reminders, and the Syft-powered analytics layer. Sage Copilot in Sage 50 v31.1 (the March 2026 update) ships invoicing, reporting, document capture, and AI fraud detection. BrightPay's 2026 marketing post points at anomaly detection and intelligent capture as the next direction. These are real shipping features.

What follows is what each vendor actually ships in May 2026, where Ireland availability lags, and the lanes where a partner-built workflow still earns the price.

For the buy-vs-build framing in general we use the add AI to existing software decision tree. This post is the practice-specific version: same shape, named vendors, Irish-specific gaps.

Xero JAX: what is GA, what is beta, what is still rolling out to Ireland

Xero began rolling JAX out from September 2025 starting with US and Australian tenants. The company's own positioning says it will "expand to other regions in the coming months." The Xero UK JAX product page is live, but a pinned general-availability (GA) date for Ireland is not surfaced in any verifiable Xero source as of May 2026. UK accountants are already touching JAX in beta; Irish practices are watching the rollout list.

What JAX ships where it is available (per TechRepublic's September 2025 update note and Xero's own beta blog):

• Predict invoice payment dates
• Automate expense processing
• Progress payments (deposits and partial)
• Syft-powered analytics layer for advisor questions
• Automatic Bank Reconciliation in beta since November 2025, targeting over 80% of statement lines auto-matched

The H2 2026 roadmap announced at Xerocon Brisbane points at AI-assisted bill payments, anonymous business benchmarking, and native budgeting. None of those are shipped yet.

What JAX does not do today, anywhere: cross-system orchestration with BrightPay, Surf Accounts, ROS (Revenue Online Service), or practice-management tools; chase loops that span multiple tools; anything outside the Xero plus Syft surface.

The Irish-specific point most reviewers miss: Xero's native Revenue integration covers VAT3 only, via the free Parolla partnership. CT1 (the Irish corporation tax return), Form 11 (the Irish self-assessment income tax return), RCT (relevant contracts tax), and reverse-charge VAT are not supported natively. Vendors.ie's 2026 Xero Ireland review sets this out in detail. For a practice with sole traders, construction subcontractors, or Irish-resident corporates, the bridge to ROS is the gap JAX was never going to close.

Sage Copilot: GA in the UK, marketing in Ireland, pricing not yet localised

Sage Copilot is GA in the UK in Sage 50 Accounts v31.1+ (the March 2026 update) and across Sage for Accountants. Ireland presence is currently a marketing one: sage.com/en-ie/blog/sage-copilot is live, but a separate Ireland GA date and Ireland-specific pricing SKU are not published as of May 2026.

UK pricing per Finexer's summary: one Sage Copilot seat is bundled with every Sage Accounting plan, and additional Copilot users cost roughly £20 per user per month. The Sage Accounting subscription itself starts at £18 per month plus VAT, with £0 for the first three months on the Start tier. A specific Irish Copilot SKU price is not published as of May 2026.

What Sage 50 v31.1 actually shipped in March 2026:

• Natural-language Report Search
• AI Document Capture
• Automated Purchase Processing
• Receipt Automation (OCR, or optical character recognition)
• AI fraud detection
• AI-powered bank feeds

The Intacct side of the Sage estate is further ahead. Per CPA Practice Advisor's coverage and the Rand Group release-2 notes: Close Assistant is GA in UK and US, Subledger Reconciliation Assistant is in early adopter, Variance Analysis (gen-AI explanations) is part of Close Automation, and a Finance Intelligence agent went into early access in December 2025. Practices serving mid-market clients on Intacct are dealing with a more mature AI stack than the Xero or Sage 50 one.

BrightPay: AI sits on the 2026 roadmap with no GA features yet

This one matters because it is the cleanest gap in the Irish stack.

BrightPay's 2026 innovation post describes anomaly detection, predictive payroll cost forecasting, and intelligent document capture as their next frontier. The wording is explicit: these are roadmap items rather than shipping product. The cloud version (BrightPay Cloud) launched January 2025 and BrightPay Connect provides employer portals and employee self-service. The 2026-27 product priority per BrightPay's own availability docs is closing the desktop-to-cloud feature gap, particularly around the My Future Fund auto-enrolment scheme that launched September 2025. The work shipping in payroll is real engineering work on auto-enrolment and desktop-to-cloud parity, and it sits entirely outside the AI roadmap.

For a practice running BrightPay for clients' payroll (which is most Irish practices), there is no vendor AI on this surface in May 2026.

The Bright group as a whole: zero GA AI features

This is the single biggest Ireland-specific finding in the article. Bright Software Group was formed when Thesaurus and Relate merged in September 2021, then expanded with AccountancyManager in March 2022 and BTCSoftware in October 2022. The group currently ships:

• BrightPay (payroll): no GA AI
• BrightPay Connect (portal): no AI
• Surf Accounts and Surf Accounts Production (ledger, Irish-resident, Azure-Dublin hosted, multi-currency, VAT3/RTD/VIES/INTRASTAT support): no AI features marketed on the accountant page, the Relate Software product page, or the CPA Ireland Software Showcase listing
• BrightManager (practice management, formerly AccountancyManager): template-and-rules engine for onboarding, AML (anti-money laundering) letters, and deadline tracking
• BrightTax (formerly BTCSoftware): MTD-for-IT (Making Tax Digital for Income Tax) ready compliance, rule-based, no AI claims
• Thesaurus (legacy desktop payroll for Ireland): no AI

A practice running the full Bright stack has zero vendor AI in May 2026. That is the cleanest partner-built lane on the entire Irish estate, because for these practices the alternative to a partner build is no AI at all until Bright ships something.

A correction worth making in print: Big Red Cloud is not part of Bright

Common misconception in the Irish market. Big Red Cloud / Big Red Book is a separate Irish company. The Bright group has no stake in it. They are also further ahead on AI than any other Irish-domestic vendor: their Purchase Importer is GA, powered by Google AI, and extracts supplier, totals, VAT, and dates from photo, email, or upload. Their own marketing claims around 3 million invoices a month processed through it. Big Red Book (the legacy desktop product) is sunsetting; vendor support ends January 2027 and those customers are being pushed to the cloud product. SME tier pricing starts at around 399 EUR per year.

OmniPro / OmniHUB: content and workflow tools with no AI layer

For completeness on the practice-management side: OmniPro's practice management product and OmniHUB Ultimate 2025 ship compliance templates, CA14, FRS 102 (the Irish/UK GAAP standard), AML tooling, and a 750-plus question KnowledgeHUB. It is a strong content and workflow product. No AI layer.

Dext and AutoEntry: where the bookkeeping-extraction layer actually lives

Two corrections up front, because the Irish market gets both wrong often.

AutoEntry belongs to Sage. A persistent Irish-market belief is that it sits inside the Bright group; it does not. Sage acquired AutoEntry in 2019, confirmed in the company's PitchBook record and noted on AutoEntry's own news page. It ships AI-OCR with full line-item extraction, multi-language support, supplier-level rules, and integrations with QBO, Xero, Sage, FreeAgent, KashFlow, SortMyBooks, and ClearBooks. Credit-based pricing on the entry tier sits in the high-teens EUR per month range. Verify the live credit allowance on AutoEntry's pricing page on the day you sign.

Dext AI Assist launched March 2026. It sits on top of the OCR (optical character recognition) layer and learns supplier-level categorisation patterns, suggesting tax treatment per supplier rather than per document. Dext's own marketing line on capture is 99% plus, which is a top-of-band claim from a controlled corpus rather than a field result. Independent OCR benchmarks paint a more honest picture. Zerentry's 2026 200-document field-level test put Dext at 75 to 95%, AutoEntry at 73 to 93%, and the now-discontinued Hubdoc at 65 to 90%. Treat the vendor headline as the ceiling. The field average sits below it.

The bigger event for any Xero-using Irish practice is the Hubdoc shutdown. Hubdoc retires on 8 May 2026 and is being replaced by Xero Files. Xero Files is storage only: no line-item extraction, no auto-fetch from supplier portals, no rules engine. Xero Central's own help article confirms the change. Every Xero-using practice in Ireland is being forced to make a receipt-capture decision this month: Dext, AutoEntry, or a partner-built pipeline. The cost difference between the three at 250 clients is the question we walk through later in the TCO table.

QuickBooks Intuit Assist: US GA, UK beta, Ireland not announced

A short paragraph because some Irish practices serve QBO-based clients. Intuit Assist for QuickBooks is GA in the US for QBO web and mobile. Per Intuit's investor release, specialised agents are shipping: an Accounting Agent (categorisation, reconciliation), a Payments Agent (accounts receivable, payment-date prediction, collections), a Customer Management Agent, and a Finance Operations Agent. The UK press page confirms UK beta for "selected QuickBooks customers who register their interest." Ireland: no separate launch announcement. Pricing is bundled into the QBO subscription.

Microsoft 365 Copilot: the compliance landmine on every tenant

Most practices already pay for Microsoft 365 and treat Copilot as the cheapest AI on the estate. There is a buried compliance switch worth pulling first.

Flex Routing went default-on for tenants created after 25 March 2026. It sends inference traffic outside the EU during peak demand unless the tenant opts back to EU-only. For a practice handling client financial data under GDPR and the EU AI Act, the IT lead has to opt the tenant back to "EU Data Boundary only" or accept that some inference goes through US data centres. We covered the AI Act residency framing in our piece on Article 99(6) and the SME fines tier. If your tenant is post-March 2026, check the routing setting today.

When the vendor wins (the honest list)

Four shapes of practice where the vendor genuinely wins and the partner build would be over-engineered:

1. Single-stack Xero practice with Dext bolted on. JAX (when Ireland GA arrives) plus Dext AI Assist will handle 80% or more of the bookkeeping motion. JAX-powered Bank Reconciliation already targets over 80% of statement lines in beta. A practice with under 50 SME clients on one core ledger will not see ROI on a partner build.
2. Single-stack QBO practice in the US, or UK practice on QBO in beta. Intuit Assist's four agents cover the same surface JAX does on Xero. Wait for the rollout.
3. Sage 50 v31.1 shop with AutoEntry already in place. Sage now owns both the ERP and the OCR layer. Copilot covers natural-language reporting, document capture, and fraud detection. A partner build here is reinventing what is already integrated.
4. Mid-market client base on Sage Intacct. Close Assistant is GA UK and US, Subledger Reconciliation is in early adopter, Finance Intelligence agent went into early access in December 2025. These are deep, audit-aware AI agents that a partner cannot rebuild in scope or in time.

If you are in one of these four shapes, do not pay for a partner build. Use what is in the box.

When the partner build wins (also honest)

Six lanes where a partner build still earns its keep at a knowable price:

1. Cross-system orchestration across Xero or QBO + BrightPay + ROS + Surf + practice management. No vendor wraps all four. JAX cannot see payroll in BrightPay; Sage Copilot cannot see Surf; BrightPay cannot see the ledger; the Bright suite has no GA AI layer. A practice running mixed clients on different stacks has no vendor that spans them. This is the strongest single wedge.
2. Multi-deadline chase loops keyed to the practice-management calendar. Vendor SaaS does not ship multi-client, multi-deadline, escalation-aware chasing tied to your own deadline tracker. Hubdoc was the closest tool and it retires on 8 May 2026. Xero Files is storage only.
3. CT1, Form 11, RCT, and reverse-charge VAT bridges to ROS. Xero only covers VAT3 via Parolla. For practices serving Irish construction (RCT), corporates (CT1), or sole traders (Form 11), the missing bridge is real. BrightTax handles compliance with a rule engine and ships no AI augmentation today.
4. EU-only data residency for AI inference. Xero JAX, Sage Copilot, and Intuit Assist all run on US-hosted models (Azure OpenAI, AWS Bedrock). A practice with regulated clients (healthcare, financial services, public sector) who need EU-residency assurances under the AI Act and GDPR cannot get them from vendor AI today. A partner-built EU-only stack (Mistral on EU infrastructure, or Gemini on europe-west4) is a genuine differentiator and a defensible procurement story.
5. Bespoke portals layered on live ledger data. AccountancyManager and BrightManager portals are template driven. A portal that surfaces live ledger, payroll, VAT-deadline, and chase status in one screen is a partner build.
6. The Bright stack practice. Bright ships zero GA AI features today. A partner-built AI layer on top of the Bright stack has an open field, and the Bright stack is one of the largest cohorts of Irish practices.

If your practice is in one of these six lanes, a partner build is the only way to get the work done in May 2026.

A worked TCO comparison: 20-staff Irish practice, 250 clients, Xero-leaning

Three columns, three years, broadly the practice shape we see most often: 20 staff, 250 clients, mostly Xero-based with BrightPay payroll, single-entity, Ireland-only. Figures are directional and should be reverified against current Xero IE and Sage IE pricing pages on the day you decide.

Cost line	Pure Xero JAX (when GA in IE)	Sage 50 v31.1 + Sage Copilot + Dext	Partner-built workflow (the /accounting-chase shape)
Software / build	Xero Premium ~78 EUR/mo x 36 = ~2,800 EUR; JAX bundled when GA	Sage 50 Plus ~140 GBP/mo + Dext per-client	14,500 EUR one-time
Implementation	minimal (already on Xero)	~3,000 EUR reconfigure and train	0 (covered by build)
Per-client AI add-on	None visible (JAX bundled)	Dext ~19 GBP/mo x 50% of book x 36	None
Maintenance / support	bundled	bundled	~1,500 EUR/yr x 3 = 4,500 EUR
Chase-admin staff time saved	~25% of chase	~40% of chase	~75% of chase (the 400+ hours/yr LP claim)
Three-year total	~2,800 EUR software + ~90k EUR unreduced chase labour	~20,500 EUR software + ~55k EUR chase labour	~19,000 EUR build and maintenance + ~30k EUR residual chase labour

The TCO numbers are directional. The pattern that matters: at 200-plus hours of chase saved per year, multiplied by loaded staff cost, the partner build pays back inside 12 to 18 months for a practice this shape. The real question is which lane your practice sits in. If your shape fits one of the six partner-wins lanes, the build pays. If your shape fits one of the four vendor-wins shapes, the SaaS pays.

We walk through the same shape of buy-vs-build decision in our AP automation post, at a different volume band.

Three call-outs where we would tell a practice not to build

The credibility move. Three Irish practice shapes where we would explicitly recommend the vendor over the partner build:

• Five-staff Irish practice, all on Xero, single entity. Wait for JAX to land in Ireland. In the meantime use Dext for receipts and a tight chase process. Do not pay 14,500 EUR.
• Sage 50 v31.1 shop with a single-entity client base. Sage Copilot plus AutoEntry covers invoicing, reporting, AI fraud detection, and OCR on day one.
• Big Red Cloud shop with high invoice volume. Their Google-AI Purchase Importer is GA and processing claimed millions of invoices a month. Do not replace it; bolt around it where you need to.

Vendor AI in May 2026 is good enough at narrow tasks that it would be dishonest to pitch a partner build at every practice. The four shapes above are the ones where buying is the right call.

Where the partner build pays back

If your practice does not fit the vendor-wins lane, here is what the partner build looks like in concrete terms: the chase loop, calendar-tracked, cross-system across Xero or Sage or Bright plus BrightPay plus ROS plus your practice-management tool, EU-residency-pinned at the inference layer, with a documented refund clause if it does not save 400-plus hours a year. Fixed price, four to eight weeks, owned outright at the end.

The shape of the build is set out at /accounting-chase. If you want a 45-minute call with Karl to run the six partner-wins lanes against your actual practice and tell you which one you are in (or whether you should wait for JAX), that is the same page.

Vendor AI in May 2026 ships real work, and it leaves real gaps. Map your practice against the six partner-wins lanes and the four vendor-wins shapes. The answer falls out.