Bill McDermott just told The Wall Street Journal he's rebuilding ServiceNow's entire business model around AI. That's a $15 billion company pivoting publicly...

If you run a software company or a BPO, this matters. Not because you should copy ServiceNow, but because they've just validated what you've probably been sensing for a while: the market is moving to digital labour and outcome-based pricing. That's not speculation anymore. It's strategy at the highest level of enterprise software.

The question isn't whether this shift is real. It's how you make the same shift without creating new dependencies you'll regret in three years.

The per-seat model is under pressure

Half of ServiceNow's new business revenue now comes from non-seat pricing. One of the largest enterprise software companies in the world has moved its commercial model away from headcount in the space of a few years.

Wall Street rewarded them for it. The market is telling us something: digital labour revenue is the growth story. If your pricing still scales with how many people your clients employ, that model is being challenged by every AI announcement, every automation deployment, and every CFO scrutinising their cost base.

ServiceNow has made their bet. The infrastructure they're building will lock clients deeper into their ecosystem. That's their business model. But if you're a BPO or software company looking to make this same shift, you need to ask: do I want my automation infrastructure tied to someone else's roadmap?

Three problems most organisations haven't solved

The shift to digital labour and outcome-based pricing is real. But most organisations trying to make that shift are walking into three problems they haven't fully thought through.

1. Independence

ServiceNow is an ecosystem play. So is Microsoft with Power Automate and Copilot. So are most of the enterprise automation platforms.

If you're a BPO or software company, your automation infrastructure is a strategic asset. It's what allows you to deliver outcomes at scale, to price services differently, to build margin into work that used to be purely labour-arbitraged. Do you want that asset tied to a single vendor's roadmap, their pricing changes, their strategic priorities?

The organisations thinking clearly about this want architectural ownership. They want to bring their own LLM, not be locked into one provider. They want to connect across their full technology stack, including the legacy systems that don't have APIs. They want to build automation they can monetise, not automation that makes them more dependent on someone else's platform.

2. Control

Pure agentic AI is impressive when it works. When it doesn't, good luck explaining to a client what happened and why.

The promise of agentic automation is autonomy: AI that reasons through problems, makes decisions, takes action. The reality is that autonomy and predictability are in tension. The more autonomous the system, the harder it is to guarantee consistent behaviour, to audit what happened, to explain why a particular decision was made.

For BPOs delivering services under contract, for software companies embedding automation in their products, for any organisation operating in a regulated environment, this matters. You need automation that behaves predictably. You need governance you can actually stand behind. You need to be able to tell a client exactly what your Digital Worker Bot did and why.

That's not a limitation of AI. It's a requirement of running a serious business.

3. The token cost problem nobody's talking about

This is the one that's going to bite people.

Pure agentic platforms burn tokens on every reasoning step, every decision, every document processed. The more autonomous the system, the more tokens it consumes. And token consumption scales with volume in ways that are genuinely difficult to predict until you're running in production.

If you're selling hours, unpredictable costs are manageable. You bill for time, costs are what they are, margin is margin. But if you're selling outcomes, if you're pricing a fixed fee for a process delivered, you need to know your cost of delivery before you sign the contract.

Most organisations haven't modelled what their agentic AI will actually cost at production volume. They've run pilots, seen impressive results, and assumed the economics will work out. They're going to find out the hard way that "agentic" and "predictable unit economics" don't naturally go together.

This isn't about being cheap. It's about being able to price outcomes with confidence.

DWIGHT Studio was built for this

We built DWIGHT Studio because the platforms available were forcing organisations into trade-offs they shouldn't have to make. Flexibility versus predictability. Autonomy versus governance. Innovation versus cost control.

DWIGHT Studio is an Automation Operating System that solves all three problems.

Independence. LLM-agnostic by design. Bring your own model, commercial or open source, and change it whenever you need to. Connect across modern and legacy systems, with or without APIs. Build automation you own and can monetise.

Control. A hybrid execution model where RPA handles deterministic tasks and AI is invoked only where it adds genuine value. Every action logged, every decision auditable, every workflow observable. Digital Worker Bots that don't hallucinate, don't behave unpredictably, and don't execute outside the boundaries you set.

Predictable economics. Because AI isn't running constantly, because you control when and how it's invoked, you can model your costs before you price your outcomes. Fixed monthly platform subscription. No token surprises at scale.

The three-wave model

DWIGHT Studio structures automation into three progressive waves. You move through them at your own pace, based on what each use case actually requires.

Wave 1: Plain RPA. Deterministic, rule-based execution. No AI involvement. Maximum predictability and the lowest cost per transaction. This is where most processes should start, because most processes don't need AI to run reliably.

Wave 2: AI-infused workflows. RPA execution enhanced with AI for specific cognitive tasks: classification, summarisation, document processing, decision support. You control when AI gets invoked and how much it costs.

Wave 3: Pure agentic automation. End-to-end AI-driven workflows with higher autonomy. Available when the use case justifies it, governed so you maintain control.

The point isn't that Wave 3 is better than Wave 1. The point is that you choose the right approach for each process, rather than having a platform push you toward maximum AI consumption because that's how their economics work.

What this looks like in practice

We've deployed this model across recruitment, healthcare, and professional services. The outcomes are measured in operational impact and known economics.

GHR Healthcare processes over 1,000 invoices per week automatically. Two full working days of manual effort freed. An extra day of DSO captured. The automation runs overnight, every night, at a predictable cost.

DeWinter Group cut contractor onboarding from 45 minutes to 4 minutes. Three team members moved from data entry to client-facing work. Total automation cost: a fraction of one FTE.

PRN Healthcare automated credential management end-to-end. Licences and certifications tracked 24/7. Reminder emails sent automatically. One FTE redeployed to relationship work.

These are production systems with known unit economics, not pilots with hopeful projections.

DWIGHT Studio gives you the architecture to build and deploy digital labour on your terms. Control over how your automation behaves. Economics you can model before you price.

The shift is happening. How you make it is up to you.