So SaaS is dead, they say. I'm still paying for mine.

SaaS plus is a model where an operator buys a core platform as its system of record and has features built on top of it, by the vendor's development team or by the operator's own AI. SaaS itself is not dead in 2026: Gartner forecasts worldwide software spending of US$1.46 trillion, up 15.2% on 2025.

I run a software company. We write code with AI every day, and every month I still pay for Xero, Slack, a CRM and a list of other subscriptions I'd rather not add up. Nobody on the team has offered to build us an accounting package. I'd be worried if they had.

My LinkedIn feed says that arrangement is finished. SaaS is dead, anyone can build anything in a weekend, and a subscription is for people who haven't heard the news.

Larger operators were wrestling with buy or build long before AI turned up. WOTSO tried both answers before it came to us: a custom build from an app company, then an off-the-shelf product. Neither worked (case study).

WOTSO is an investor in Hamlet as well as a customer, so I'm not a neutral witness. Look at where it landed all the same. It runs on a platform somebody else maintains, with its own work built on top. I think that's the right answer for most operators with more than a few sites, and AI has made it much faster to do. We've started calling it the SaaS plus model.

One caveat before the numbers. Nobody knows what software looks like in three years, me included, and with AI moving at this pace 12 months is already a stretch. So read this as a view for the next 12 months.

Is SaaS dead?

SaaS spending is still growing in 2026, even though the share prices of the companies that sell it took a beating early in the year.

In February 2026, software and services shares lost about US$830 billion in market value over six trading days after Anthropic released plug-ins for its Claude Cowork agent, according to Reuters figures reported by ITPro. Most of the "SaaS is dead" posts trace back to that week. Spending went the other way.

The surveys are right that building your own has become a real option. Look at what Retool's builders replaced, though: workflow automations, internal admin tools and reporting tools topped the list. Those are jobs around the edge of a business.

Klarna is the example everyone remembers wrongly. In 2024 its CEO told investors the company had shut down Salesforce and would shut down Workday, and the story became "AI replaces the CRM". In March 2025 he corrected it himself: "we did not replace SaaS with an LLM" (Diginomica). Klarna had moved to other SaaS products and joined up its own data underneath them.

What is going is SaaS as it used to be sold: a login per person and a bill that climbs with headcount. ServiceNow's CEO told analysts in July 2026 that about half of the company's net new business is no longer sold by the seat. Charging per person makes less sense every month that software does more of the work with nobody logged in. For an operator that's good news, because a bill that grows with your team or your member count charges you for filling the building.

Can you build your own coworking software with AI?

Yes, you can build a working booking page or a member list with AI in a weekend, and plenty of operators will. Building is now the cheap part.

The work is in everything the weekend version doesn't know about yet. Follow one member through an ordinary few months. Her card fails on the 1st, so something has to retry it and then send the reminder nobody enjoys writing. She pauses her membership over summer, and the invoices have to stop on the right day. When she pays, the payment has to show up as paid in Xero without anyone typing it in. She books the boardroom at the member rate, her client books it at the guest rate at 9pm without an account, and the GST has to be right on both.

Each of those lines is a decision somebody has to make and then keep correct when the tax rules, the payment gateway or the accounting software changes. A coworking platform is hundreds of those decisions already made and maintained. If we started Hamlet again tomorrow, AI would write the code in a fraction of the time. Knowing what to build, and how each feature should behave when real members and real money are involved, has taken us since 2018.

Retool's survey points the same way. Of the builders who had shipped software with AI, 31% prompted their way to a complete app. Most used AI to write pieces of code that they tested and assembled themselves, and the most common reason a build never reached production was a lack of technical knowledge.

What does software you built cost to run?

Software you built costs you a team. Somebody has to own it after launch, and at any real scale that means people on your payroll whose whole job is the software.

Security comes first, because a coworking system takes payments and holds members' personal details. On 29 October 2026, Veracode reported that code from 18 AI models passed its security tests 62% of the time when the prompt didn't ask for secure code, and the best models passed 73%. That's up from 56% in July, so it's improving, and it still leaves plenty for a reviewer to catch. If nobody on your team can review it, nobody is reviewing it.

Then there's upkeep. Tax rules change, the payment gateway updates its API, Xero changes something, and each one lands as a job for whoever built your system. If the whole thing lives in one person's head, you can't easily train a new starter on it or open a second site with it, and you're exposed the day that person leaves or gets busy.

Agents need the same care. Setting up one agent for invoices, another for enquiries and another for the Monday report is the current fashion, and it's the new spreadsheet chaos: every agent is one more thing to check. In June 2025, Gartner predicted that more than 40% of agentic AI projects will be cancelled by the end of 2027 because of cost, unclear value or weak risk controls. The reason to automate was to spend less time at a screen and more time in your space, with the people who decide whether it fills and stays full. A dozen agents to babysit takes you the other way.

One more number, with a caveat. MIT's NANDA project reported in August 2025 that AI tools bought from specialist vendors succeeded about 67% of the time, and internal builds about a third as often (Fortune). The sample was small and the method has been argued over, so take it as a direction.

When does building in-house make sense?

Building in-house makes sense when software is part of what sets you apart, and when you can hire and keep two or three senior people who are good at running AI.

AI is changing the hands-on work in development, testing, project management and support. In our team it has made more of us builders and given us more time with clients. The work still needs good people. Somebody decides what's worth building, teaches the AI how your business runs, checks and corrects what it produces, and answers for it when it's wrong. Those people are senior, and they're hard to hire.

So the fair comparison for 2026 and 2027 is a small senior team. Here's what that costs a month in Australia, next to the alternatives.

The working. Base salaries come from Employment Hero, Robert Half's 2026 Sydney figures and Indeed's product manager average, using the lower and higher source for each role. On top go super at the ATO's 12% and NSW payroll tax at 5.45%. The rest are my estimates: leave loading, A$5,000 a person a year for equipment and software, and a recruitment fee of 15% of salary spread over three years. Agency day rates are Cipher Projects' July 2026 figures for Sydney.

The last row of that table needs explaining. The subscription buys the standard platform, with standard support by email and in-platform help, and plenty of operators run on that alone. An in-house team does more: it builds features, meets you every month and works out better ways to run your operation. To get that from us you add the partner package and custom work, and for a large operator those usually come to more than the subscription itself. Put that total next to the salaries when you compare.

Those are the salaries. Three things sit outside the table.

Cover and time come first. A team of three has to build, test, release and stay on call for a system that takes payments, and it starts from nothing, or from a licence it can't change. Hamlet has been in build since 2018, and billing has run on it every day since the first site went live.

Then there's where the ideas come from. Good features rarely come from one person or one company. They come out of conversations between people who know the industry, people who know software, and people who run operations and finance. The core of Hamlet is shaped by many operators and paid for by subscriptions, so every operator on it gets those ideas without funding them one by one. The subscription pays for the roadmap we set. A say in what gets built and when, and a team that knows your operation, come with the partner package and custom work.

Custom work runs the other way. What a partner pays for custom work is spent on that partner's list and nobody else's. On our enterprise partner model that comes with named people on your account: a support team that knows your set-up, an account manager, and Streams reserved for your list, directed by the people who designed and built the platform.

If software is part of what makes you different and you want to own it outright, hire the team and build. I mean that. An operator with an unusual model and its own product people can build things no vendor will ever prioritise.

If your people are already building with AI, that's worth encouraging. Give each build an owner, and make sure somebody with enough technical knowledge has looked at the security, the tech stack and who updates it when things change.

There are cases where Hamlet is the wrong answer too. A single site with a handful of members will spend less on a simple booking tool and Xero. And if your model sits a long way from memberships, bookings and billing, no platform will fit well, and building may be the better call.

What is the SaaS plus model?

The SaaS plus model is buying a core platform as your system of record and having your own features built on top of it. It comes in two forms, and plenty of operators will use both.

For a CEO or COO, the first reasons are speed and fit. You can think of a feature this month and have it built at a sensible price, where a large vendor would file it under a stack of other customers' tickets. And when a platform can't do something you need, the usual fix is to bolt on another product and then pay someone to manage the join. Building the feature into the core means there's no join to manage.

The bigger reason is your data, and I think it matters more every month. An AI agent can only act on what it can see. With the member, the invoice, the booking and the door access in one system, an agent gets the whole picture in one place. Spread those across several products and the agent has to fetch from each one and match the records up, which uses more tokens, runs slower and goes wrong more often. It's also several sets of permissions to secure where there could be one. The more an operator consolidates, the better its AI will work.

WOTSO is the clearest example I know. The ASX-listed operator, now with more than 45 locations, has run on Hamlet since 2018, after trying a custom build and an off-the-shelf product first. Over that time the platform was extended to cover the storage, parking and health spaces WOTSO runs alongside its coworking. Billing went from a six-person job to a two-person job, with accuracy around 99%, and new sites now come on in batches of three or four (case study). When WOTSO's team wants something the platform doesn't do, it's scoped as custom work and built into the system they already run. Nobody goes shopping for another product.

On Hamlet, the two forms look like this.

SaaS plus custom is Hamlet Custom. You add a request to a shared board, and each piece of work is run end to end: product management, design, development, testing and release. We reply within one business day and size and quote it within two (Australian business days). The people who designed and built Hamlet scope it, the team that works in that code every day builds it, and it stays on the platform, so it's maintained and upgraded with everything else. If it's already on our roadmap, we tell you before you spend anything.

SaaS plus your own AI is the Hamlet API and the Hamlet MCP server. The API is included on every plan. The MCP server is in alpha with a small group of operators. It connects your own AI assistant, such as Claude or ChatGPT, to your Hamlet workspace, so your team can ask questions of your data, ask for a report or have the assistant take basic actions. Personal data is tokenised and every call is audited. We're releasing it slowly on purpose, because an agent that can touch member records and money should be boring before it's clever. If you'd like to try it, get in touch.

We've done custom work for a few clients for years, and AI is what lets us offer it to more operators without taking people off the partners we already have. It has changed who builds, too. My background is design and project management, and I now ship code to Hamlet myself, because I'm the one closest to what customers are asking for. It goes through the same review, testing and release as everyone else's work before it reaches a customer. It has also let us go back and improve things we never had the capacity for, and build automations and AI workflows we couldn't have built before.

None of this changes the rule we work to. Software takes the jobs where a person adds nothing: raising invoices, retrying a failed payment, sending the receipt. The jobs that need a person stay with a person: the member who wants to cancel, the late payer with a reason, somebody's first morning in your space.

What changes in the next 12 months?

Over the next 12 months, building software will keep getting easier and faster, and pricing will keep moving away from the seat. Past that, I don't know, and I'd be careful of anyone who says they do.

Here's what I'd put money on.

Building gets easier again. Veracode's security pass rate for AI-written code moved from 56% to 62% in three months, after a year of going nowhere. The tools we use to build Hamlet today will look clumsy by this time next year. Fewer people will write code, test it and answer tickets by hand, and more will spend their day directing AI and checking what it did.

I'd also expect more vendors to charge for what the software does, and fewer to charge for how many people log in. Hamlet prices by location with unlimited members, and our own AI features will run on usage credits when they launch.

The third bet is that software opens up to your AI. I think most serious platforms will let your own assistant read and act on your data within a year or so, the way our MCP server does in alpha now. A system that can't will start to feel like a filing cabinet with a login.

And here's what I can't tell you. By 2028 an operator might describe the system they want over a coffee and have it running by lunch. It could happen. Even then, somebody has to know whether it works, whether it's built right for the business and whether it's still secure in month 13. That part doesn't get automated away quickly, and it needs an owner.

So decide for a year. Whatever you buy, check you can get your data out, that it has an API, and that you aren't locked in for long. Hamlet's monthly plans run on 30 days' notice, for what it's worth.

What should you buy, and what should you build?

Buy the system that holds your money and your member records, build on top of it where your space is different, and build the small stuff yourself.

You can run this on your own business this week. Ask finance for every software subscription, then ask each site for every tool, spreadsheet and agent somebody has built. Next to each one, answer three questions.

1. Does it hold money or member records?
2. If it breaks on a Saturday, who fixes it?
3. Could a new starter use it on their first day?

Anything that holds money or member records belongs in something you buy. Then count how many of the answers to question two are one person's name.

If you run several sites and there's a list of things your current system can't do, send us the list. We reply within one business day, and you'll be talking to the people who would build it.

Questions operators ask

Is per-seat software pricing going away?

Per-seat pricing is losing ground as software does more work with nobody logged in. ServiceNow said in July 2026 that about half its net new business is no longer sold by the seat. Hamlet doesn't charge per member: it's priced by location with unlimited members, from A$350 a month for one location, excluding GST, as of November 2026.

How much does it cost to build coworking software in-house?

Building coworking software in-house in 2026 takes a small senior team that can direct AI, plus a platform to build on. A senior developer, a second developer and one person covering product and testing cost about A$40,000 to A$45,000 a month in Australia once super, payroll tax, equipment and recruitment are counted, on 2026 salary data from Employment Hero, Robert Half and Indeed. The like-for-like comparison on Hamlet is the subscription plus a partner package and custom work, which are priced with you. The subscription alone, from A$2,595 a month for ten locations excluding GST, covers the standard platform and standard support and does not include a team building for you.

What is the SaaS plus model?

The SaaS plus model is a way of buying software where an operator subscribes to a core platform as its system of record and has features built on top of it. The vendor's development team can build them (SaaS plus custom), or the operator's own team can build them with AI through the platform's API or MCP server. On Hamlet, the two forms are Hamlet Custom and the Hamlet MCP server.

Why does keeping data in one system matter for AI?

An AI agent can only act on what it can see. With members, invoices, bookings and access in one system, an agent gets the full picture in one place, uses fewer tokens, and works faster and more reliably than one matching records across several products. One system also means one set of permissions to secure.

Can I connect my own AI assistant to Hamlet?

Yes. The Hamlet MCP server connects your own AI assistant, such as Claude or ChatGPT, to your Hamlet workspace, with personal data tokenised and every call audited. It's in alpha with pilot operators as of November 2026. AI features built into Hamlet itself haven't launched yet.

Can Hamlet build a custom feature or integration for us?

Yes, through Hamlet Custom. The team that builds Hamlet builds custom integrations, automations, reports and portal changes on your workspace. Hamlet replies to a request within one business day and sizes and quotes it within two (Australian business days), and bugs and support are never billed.

Automate the admin, grow the relationship.