OpenClaw 2.0 shipped the morning of this interview, and Xero, an accounting platform serving about five million small businesses, has not allowed a single employee to run it.
Most software chief executives asked about a new agent tool describe a pilot. Sukhinder Singh Cassidy described a prohibition, and gave the reason in one line: her company processes financial transactions and holds customer data.
"So today we do not let user our employees use OpenClaw environment."
Cassidy runs Xero, has been an entrepreneur three times over, has spent years as a venture investor, and was on Salesforce's earnings call the previous night talking about Agentforce — so she is answering as both a seller of AI software and a buyer of it.
I listened to the full segment so you can skip it.
Here are the 6 takeaways that matter.
👤 Guest: Sukhinder Singh Cassidy, chief executive of the accounting software company Xero, a three-time entrepreneur and former venture investor who was on Salesforce's earnings call the night before
🎙️ Host: Akash Pasricha, who anchors TITV, The Information's live weekday show at 10 a.m. Pacific
📰 Published: 31 August 2026 on YouTube (The Information)
🔴 YouTube | ⏱️ 11 min
Key Takeaways
Xero has never charged per seat, which takes it out of the AI pricing argument entirely
It prices per organization, and keeps consumption pricing in reserve for niche premium features
Small business customers want a bill they can predict, and that constrains what she can sell
The company bans OpenClaw for employees because of what it holds, not what the tool does
It processes financial transactions and holds customer data, so the protocols are strict
Xero buys ads inside ChatGPT and has since the early days, and still calls the returns unproven
It spends hundreds of millions of dollars on advertising, of which this is a small part
Its bookkeeping agent runs at a 96% accuracy rate across hundreds of thousands of customers
API traffic is 6 times what it was in January, which is why it built a low-code agent builder
Her answer to the SaaS apocalypse is that software is one layer of what Xero sells
The rest is bank connections, payment rails, regulation and tax filing, and none of it is coded over a weekend
1. Xero was never seat-based
Pasricha came straight out of the outcome-based pricing segment and asked whether Xero is playing with the model.
The premise did not apply. "I think for our own customers today, we've never been seat-based, believe it or not. We've been organization-based." Its flexible pricing today is tied to the payments business
AI is going in as an embedded feature rather than a line item, with consumption pricing held back. She said Xero is embedding several AI features and reserving the option to add consumption-based pricing to specific premium ones
The constraint is the customer. Small businesses want certainty about what they will pay, she said, while a niche premium feature that most customers do not want is exactly the kind of thing that suits paying by consumption
On the industry-wide question, she said the pricing model does not remove the planning problem. Whatever the model — fixed plus usage, outcome-based, resolution-based — the buyer still has to forecast the bill and weigh the value against the cost
Xero is also a buyer here: she named Salesforce and AWS among the enterprise vendors it works with
2. No OpenClaw for employees
OpenClaw 2.0 was released that morning and Pasricha asked which companies are letting it into their workflows.
Xero is not one of them. "Today we have not rolled OpenClaw inside of our organization."
She said the company keeps track of the software and tinkers with it to understand what is possible, but has not deployed it
Asked whether employees might be using it anyway, she said the controls are formal. "This is I mean we have pretty strict protocols. Again just think about what we do. We both process financial transactions. We hold data. So today we do not let user our employees use OpenClaw environment."
Internally the company runs a mix of models rather than standardizing. She named Claude as the predominant one and Glean alongside it, and described a level of model agnosticity as deliberate
3. Buying ads inside ChatGPT
Pasricha raised OpenAI saying it had reached a billion dollars in annual recurring revenue from advertising, and asked who is actually buying.
Xero is, and its marketing looks more like a consumer company's than an enterprise vendor's. "We are in fact spending on ChatGPT as well as you know, multiple ad platforms, top of funnel, mid-funnel, bottom funnel."
The spending base is large. "We spend hundreds of millions of dollars on ads"
She would not claim the channel is working yet. "I think it's hard to compare the returns on ChatGPT yet to the returns you get, let's say, on a search ad."
"Is it a substantial part yet of our marketing mix? It is not yet."
The reason to be there anyway is coverage, not measured return. She said the most important thing is being everywhere, that Xero has advertised on ChatGPT since the early days and was early into the ads beta, and that where its customers go it will go and test
4. Bank reconciliation at 96%
Asked whether Xero relies on closed models or open-weight ones, Cassidy walked up the ladder of what it ships instead.
The entry-level features are embedded and invisible, such as smarter capture of receipts and other documents
The headline feature does the bookkeeping. She described auto bank reconciliation as the platform doing the customer's bookkeeping for them, and gave its performance: "That right now is a 96% accuracy rate and it's already across hundreds of thousands of customers in terms of usage."
At the advanced end is a tool for building on top of Xero. "Xero Force is our low code no code agent for those I we'd say accountants and bookkeepers and third-party app providers in our ecosystem that want to build custom apps on top of Xero in a secure environment."
The company also announced Casper in the US, which she described as a ledger-agnostic AI assistant — meaning it works across any ledger, not only Xero's
The reason the builder tools matter is what the developers are already doing. "but right now on our APIs we have 6x the volume 6x since January"
5. A 47% drop, rule of 48
Pasricha put the market's verdict to her: the show's own figure was that Xero's stock is down about 47% over the last year, though it has crept up over the last month, and he asked whether investors are pricing the SaaS apocalypse — the fear that AI destroys the business model of software companies.
She accepted the diagnosis in one word. "Predominantly SaaS apocalypse?" She said the company hit all-time highs a year ago and then suffered alongside others in a period when investors could not separate the winners from the losers
Her rebuttal is that Xero is more than an application layer, listing infrastructure, data, compliance rails, applications and an ecosystem as separate assets
The financial answer she gave for the year announced in May. "this past year we announced in May we grew 31% we generated almost $800 million NZ in free cash flow. We're rule of 48."
Her advice to holders was to wait. "So I think look I think this favors patience." She noted people had started talking about a SaaS renaissance that week off the back of Salesforce's earnings
"I mean I think the most important thing right now is to keep executing for our customers."
On the model layer she described the company as neutral by design. She said Xero is integrated with Microsoft, OpenAI and Claude: "So really we are an open stack with a lot of proprietary assets and you know an offensive play."
6. Build your own tax filing
Pasricha closed with the argument that hangs over every software company: that a customer can now build the software themselves with AI tools.
Cassidy said the coding is the easy layer, and not the one anybody wants to own. Her reply was that applications are one layer of Xero's stack, that anyone who wants to build application software should go ahead, and that as a three-time entrepreneur the last thing she would ever have chosen to code is her own accounting system
The part that is not a weekend project is the regulated plumbing. "we're connected to thousands of banks we're connected to payment rails we're regulated we're connected to tax authorities we file your sales tax your payroll taxes"
Her offer to anyone who wants to maintain that infrastructure themselves was to knock themselves out
The strategy is to open that plumbing rather than defend it. She said Xero will keep it running privately and securely and open it to both its own first-party applications and third-party ones built on Xero
Bonus Insights
She was on Salesforce's earnings call the night before this interview, talking about Agentforce, and referred back to the pricing discussion in the segment immediately before hers
Her forecast for AI pricing generally is that several models will overlap rather than one winning: fixed plus usage of some kind, or outcome-based
She framed Xero's position as the AI provider for its own customers, whose sophistication ranges from small businesses taking whatever is embedded to accountants building custom applications on the API
"We're going to keep playing offense on AI because that's what our customers need us to do"
This was the third of five interviews in one episode; the other four are written up separately
Cassidy's bottom line is that the market is pricing Xero as an application company at risk from AI, and her answer is that the applications are the thin layer — the bank connections, payment rails, regulatory standing and tax filing underneath them are what a customer cannot rebuild, and Xero intends to open them rather than guard them.
Products, Companies & Tools Mentioned
Xero (Her company: accounting software for roughly five million small businesses, priced by organization rather than by seat, and down about 47% over the past year on the show's figure)
OpenClaw (Version 2.0 shipped that morning; Xero has not deployed it and does not permit employees to use it)
Claude and Glean (The tools she named as predominant internally, inside a deliberately model-agnostic mix)
ChatGPT (Where Xero has bought ads since the early days and was in the beta, though she says the returns are not yet comparable to a search ad)
OpenAI (Reported that morning to have reached a billion dollars in annual recurring advertising revenue, which prompted the advertising question)
Salesforce and AWS (Named as enterprise vendors Xero buys from; she was on Salesforce's earnings call the previous night discussing Agentforce)
Microsoft (One of the three platforms she says Xero is integrated with, alongside OpenAI and Claude)
Books & Resources Mentioned
Salesforce Is Overhauling the Way It Charges for AI (The Information's column on the pricing shift, discussed in the segment immediately before hers and referred back to in her first answer)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: