This Week in SaaS: Accounting Platforms Open Up to AI Agents, Google’s Spam Update Bites, and a Zero-Day Hits FortiMail
AI is now plumbing, not a feature: finance vendors are wiring their ledgers into assistants like Claude and ChatGPT, while attackers are using the same technology to move faster than most patch cycles. Google’s September spam update is still shaking rankings, and a critical Fortinet zero-day is being exploited right now. Here is what founders and IT managers need to know in SaaS news this week.
💳 Finance & Invoicing
Acumatica 2026 R2 Adds a Data Warehouse and MCP Support
Acumatica’s October 1 release introduces a native data warehouse that separates reporting queries from the live transactional system, so heavy reports stop slowing down day-to-day bookkeeping. It also adds Model Context Protocol support, letting you connect assistants such as Claude or ChatGPT while keeping role-based security intact. New warnings flag when AI unit usage hits a threshold, which is a sensible answer to surprise bills. The catch: the Classic UI is deprecated and classic dashboards disappear in 2027 R1, so plan your migration now. For finance teams already running Acumatica, the practical takeaway is to test the AI features on a sandbox tenant before letting any assistant near live ledgers.
BILL and Blackbaud Embed Vendor Payments Inside Financial Edge NXT
BILL announced on September 29 that its payment network will be built into Blackbaud Financial Edge NXT as Payment Assistant. Invoice approval, payment processing and reconciliation now happen in one system, with tiered approval controls and access to BILL’s network of roughly nine million businesses. Rollout starts with an early adopter program and limited US availability, and only domestic vendor payments are supported at first. It is a useful signal that accounts payable is being consolidated into the system of record rather than bolted on beside it. If you run nonprofit or education finance, ask your rep about early access, since limited rollouts tend to shape final pricing.
FreshBooks Connects to Claude Through Anthropic’s MCP Connector
FreshBooks announced on September 25 that small businesses can link their account to Claude in about 30 seconds via the connector store. The company cites a survey in which 60 percent of small business owners already use AI, mostly for marketing and customer service rather than finances. Owners decide what data the assistant can see, and the company stresses that financial decisions stay with the human. For solopreneurs, this is the lowest-friction way yet to ask plain-English questions about invoices and cash flow. Before connecting anything, review the permission scopes carefully and confirm who in your business can authorize the link.
Read more → BNN Bloomberg (press release)
📊 Marketing & SEO
Google’s September Spam Update Rolls Through Two Phases
Search Engine Roundtable reports that the September 2026 spam update arrived in waves: the first hit over the last weekend of September, and a second phase followed on September 30. Rankings have moved noticeably for affected sites. If you saw a traffic drop this week, hold off on drastic rewrites until the rollout is confirmed complete. Then audit thin, scaled or low-value pages first, because those are what spam updates typically target. Check Search Console’s performance report by page type rather than sitewide, because spam actions often hit only a section of a domain.
Read more → Search Engine Roundtable
Google Loyalty Customer Match Expands to AI Mode and 14 Countries
Google’s merchant loyalty documentation now covers 14 countries, adding India, Italy, the Netherlands, South Korea and Spain. The docs now mention AI Mode and Gemini, and the feature is reachable through the Merchant API. Participation is explicitly open to non-advertisers, and programs cannot require eventual payment, including certain free trials. For ecommerce teams, loyalty data is becoming a lever inside AI-driven shopping surfaces, not just ads. Smaller shops should read the new eligibility rules closely before building a program around them.
Read more → Search Engine Roundtable
That AI Overviews Tracking Parameter Was Not Google
Site owners spotted an ?st_source=ai_overview parameter on links in AI Overviews and hoped it meant proper attribution reporting was coming. A correction shows it was added by a Chrome browser extension, apparently to collect data for sale, not by Google. The lesson for analytics teams is practical: do not build dashboards on unofficial parameters, and treat mystery URL tags as a possible privacy issue. Real AI Overview attribution in Search Console is still something Google has not delivered. If you sell data-driven services, audit which browser extensions your own team has installed.
Read more → Search Engine Roundtable
🔐 Cybersecurity & VPN
Fortinet Warns of Critical FortiMail Flaw Exploited as a Zero-Day
CVE-2026-104286 is a path traversal and null byte injection bug in FortiMail’s management interface, scored 9.8, that lets an unauthenticated attacker write arbitrary files. It affects FortiMail 8.0.0–8.0.1, 7.6.0–7.6.6, 7.4.0–7.4.8 and 7.2.0–7.2.9, and attacks are already underway. Patched versions 7.4.9, 7.6.7 and 8.0.2 are still forthcoming, so disable the IBE feature or restrict management access to trusted networks now. CISA set a federal mitigation deadline of October 4, a good benchmark for your own. Treat the interim workarounds as mandatory, not optional, until the patched builds actually ship.
GitLab Patches Critical RCE in Self-Hosted AI Gateway
GitLab disclosed CVE-2026-90970, a critical flaw that lets authenticated users with Duo Agent Platform access escape the prompt template sandbox through a crafted flow configuration and execute arbitrary commands. Self-hosted installations should move to 19.2.4, 19.3.2 or 19.4.1. GitLab-hosted instances are already protected. This is a reminder that AI gateways are privileged infrastructure, and sandboxing around prompts deserves the same scrutiny as any other code execution path. If you self-host GitLab with Duo features, add this to today’s patch window rather than the next monthly cycle.
Microsoft: Attackers Are Winning the Early AI Race
Microsoft’s 2026 Digital Defense Report concludes that threat actors are getting AI’s benefits first. The median gap between vulnerability discovery and weaponization is now well below 24 hours, and post-compromise activity such as data theft happens in minutes. State-backed groups already use AI for vulnerability research and social engineering, including deepfake IT-worker personas. Microsoft warns the remediation gap could produce a multi-year spike in unpatched known flaws, which makes patch speed a board-level metric. The practical response is automated patching and tighter exposure management, not more dashboards.
What to watch next week
Watch for Fortinet’s patched FortiMail builds and whether Google confirms the spam update has finished rolling out. Also keep an eye on which accounting vendor ships the next AI connector, because the MCP wave is clearly just starting.