It wasn’t a blip
A lot has happened since our last update in mid-February.
The SaaSpocalypse we flagged deepened through March before partially recovering as large-cap SaaS companies showed they could embed AI into their own products. The panic subsided. The structural question has not.
Anthropic reportedly surpassed OpenAI in annualised revenue this week - reaching $30 billion ARR against OpenAI’s $24 billion, with Claude Code alone going from zero to $2.5 billion in nine months. A company that barely registered in our dataset eighteen months ago now generates more revenue than its largest competitor.
At the Weel CFO Summit in March, our CEO Daniel Kniaz described what our engineering team is experiencing firsthand: a single engineer built a Microsoft Dynamics integration in 48 hours - a project that previously took a team an entire quarter. His framing: "Software development is dead, but software engineering is more alive than ever."
This quarter’s index covers both threads: the continued rise of Anthropic, and the first data evidence that traditional SaaS spending is starting to erode.
Snapshot of findings
1. Overall adoption hit 27.5% - the fastest quarterly acceleration on record
AI adoption among Australian SMBs reached 27.5% in March 2026, up from 22.1% in January - a 5.4 percentage point increase in two months, the largest we've recorded.
For context: it took nearly all of 2024 to move from 10% to 15%. The seasonal January dip we flagged in the previous update proved temporary, with February snapping back to 26.0% before March pushed higher again.
2. The Anthropic breakout: from 5.4% to 10.2% in two months
Last quarter we flagged Anthropic's acceleration as the story worth watching. The question was whether it represented genuine platform shifting or temporary experimentation. Two months of data have answered that.
Anthropic adoption rose from 5.4% in January to 6.6% in February to 10.2% in March - an 89% increase in two months. For perspective, Anthropic sat at 2.2% in January 2025. In twelve months, it has almost quintupled.
OpenAI, by contrast, remains essentially flat at 19.6% - up marginally from 18.6% in January. The January dip we noted last quarter appears to have been seasonal noise, consistent with our earlier hypothesis. But the growth story is no longer OpenAI's.
Part of the explanation may be compositional. Anthropic's own research shows nearly half of Australian Claude usage is work-related - which likely explains why the growth registers so clearly in business transaction data.
3. It's not just trial accounts: companies spend 3x more on Anthropic
The adoption percentages tell one story. The spend per company tells the same story, more emphatically.
In March 2026, the average Anthropic-paying company on Weel spent $873 per month, compared to $286 for OpenAI. That's a 3x gap - and it's been widening. In September 2025, the equivalent figures were $399 (Anthropic) and $304 (OpenAI). Anthropic's average monthly spend has more than doubled in six months. OpenAI's has declined slightly.
More companies adopting Anthropic, each spending significantly more than their OpenAI counterparts - when multiplied out, even our microcosm of Australian SMBs mirrors the Anthropic-over-OpenAI revenue overtake playing out globally.
This likely reflects different usage profiles. OpenAI spend appears weighted toward ChatGPT seat subscriptions - relatively uniform monthly fees. Anthropic's higher average may suggest heavier API usage, Claude Code licences, and workflow automation - the kind of spend associated with operational integration rather than individual productivity.
If that interpretation holds, the Anthropic surge isn't just preference switching. It may signal a shift in how AI is being used - from productivity tool to operational infrastructure.
4. Financial Services reaches 53%; every major sector now above 20%
Financial Services has extended its lead as the top-adopting industry at 53.3% - up from 45% in January. More than half of Financial Services companies on Weel are now paying for AI.
Technology sits at 48.9% (up from 40%), Manufacturing at 39.2% (up from 34.1%).
The more striking development may be at the other end: every industry we track has now crossed 20% adoption. Construction (26.6%), Education (28.5%), Retail (26.7%), Healthcare (23.2%) - all have passed a threshold the overall index crossed in mid-2025. AI adoption is no longer concentrated in the obvious sectors.
5. A new dataset: AI spending rises as SaaS erodes
This is the new addition to the index, and it may prove the most consequential over time.
We've added a "SaaS index" - the percentage of Weel companies spending on a basket of the top 20 SaaS platforms (it’s all the obvious names minus Google and Microsoft) - alongside the AI trendline. The trajectories are diverging.
In early 2024, roughly 44% of companies were paying for top-20 SaaS tools. By March 2026, that figure has eased to 42.7%. A decline of roughly 3 percentage points over two years - gradual, but consistent.
Over the same period, AI adoption went from 10.8% to 27.5%.
The effect is more pronounced in the Technology sector, where AI adoption has risen sharply while SaaS index spending has eroded faster than the cross-industry average.
We want to be precise about what this does and doesn't show. Correlation is not causation. The SaaS decline may partly reflect vendor consolidation, pricing changes, or unrelated churn. But the coincidence - AI spend accelerating as SaaS spend declines - is now visible enough to warrant deliberate tracking.
As our CEO Dan put it at the Summit: the shift isn't that businesses are doing less. It's that they're achieving the same outcomes with fewer tools - or with AI doing what a SaaS workflow used to.
Intent vs actual: what CFOs say and what they spend
Our spending data captures what businesses are buying, but not whether it's delivering.
This week, we're also publishing our annual 2026 CFO Report, conducted in partnership with YouGov across 270 ANZ finance leaders.
The two datasets tell the same story on adoption - the CFO Report finds 54% of finance leaders are actively scaling AI, up from just 15% in 2025. However, while 84% report positive outcomes from AI or Automation, only 35% describe those benefits as clear and measurable. The remaining 49% see "positive early signs, not yet conclusive."
Professor Prabhu Sivabalan of UTS, presenting also at the Weel CFO Summit, framed this gap directly. His emerging research with LSE and Oxford on AI ROI measurement led him to conclude: "There truly is a science for measuring returns on AI investments" - but most organisations haven't built the infrastructure to apply it yet.
The commitment is clear. The evidence that it's paying off is not as clear - yet. That gap is the one thing that could slow or reverse the spending trajectory we're tracking here in the Weel AI Spend Index. Unlikely, but worth watching.
What we're watching
Platform dynamics
Anthropic's trajectory is the primary watchpoint. 10.2% in March from 3.5% in September is exceptional growth. If the rate holds, Anthropic approaches OpenAI's adoption level among Australian SMBs by late 2026. If it plateaus, the surge may reflect a one-time wave driven by Claude Code's launch rather than a structural shift. We'll be watching whether the curve sustains or flattens.
The SaaS crossover
The AI-up, SaaS-down pattern is now visible in the data. If it persists through Q2, it warrants dedicated analysis: which SaaS categories are most affected, and whether the decline reflects cancellation, consolidation, or active AI replacement.
The ROI reckoning
The CFO Report's trust gap - 84% positive outcomes, only 35% measurable - is a leading indicator for the sustainability of AI investment. Narrowing that gap likely accelerates spend further. Widening it risks a correction - not in adoption itself, but in investment scale and confidence.
Market fragmentation
Manus AI at 0.6% is marginal, but worth watching - particularly given Meta's $2 billion acquisition of Manus in early January. With Meta's distribution behind it, that number could move quickly. Combined with Perplexity at 0.8%, the market may be fragmenting beyond a two-player race sooner than expected. What the past two quarters have shown is that things move fast - and the winner is far from a foregone conclusion at this stage of the race.
Limitations (unchanged)
To maintain transparency, we reiterate the same limitations as prior releases:
- Company size representation: Data reflects Australian companies using Weel, biased toward SMB adoption patterns
- Scope: This index is a leading indicator, not a comprehensive economic measurement
- Coverage window: Analysis now spans January 2023 - March 2026
- Exclusion of bundled AI services: Microsoft Co-Pilot and Google Gemini cannot currently be isolated from broader subscription SKUs
- SaaS index definition: The SaaS index represents spending on a basket of the top 20 SaaS platforms by prevalence in the Weel dataset. Changes in the index may reflect pricing, consolidation, or churn unrelated to AI displacement.
Stay updated
This index will continue to be updated quarterly as adoption accelerates and the dataset expands. The next scheduled coverage will be in Q3 2026.
For other deeper analyses and upcoming reports from the Weel Data Research series, please subscribe to our newsletter.
