Rethinking go-to-market for the AI era

For years, enterprises have been promised that the next tool, platform or dashboard would finally unlock real marketing intelligence. Yet each wave of technology has delivered the same outcome: more screens, more noise and no clearer line of sight to commercial impact.

AI has made that model untenable. You cannot dashboard your way into an intelligent go-to-market system.

So instead of buying the latest AI add-on, the GTM team at Thoughtworks made a different choice. We rejected the tool-first mindset entirely and rebuilt their operating model around data, business logic and agents, not interfaces. We rebuilt core GTM systems; rewired for agents and reimagined value in shorter sales cycles and higher win rates.

Most companies think AI4GTM means buying a few AI-powered SaaS tools. What’s wrong with that approach?

Buying tools is the easiest way to delay the real work.

When generative AI went mainstream, everyone rushed to land on an AI strategy. For some, that meant buying whatever their existing vendors released. Salesforce announced Agentforce, Demandbase launched account summaries, Adobe added generative features. So many marketing organizations defaulted to vendor roadmaps.

The problem is that SaaS vendors don’t build for your business logic. They build for an average customer so the system can roughly fit the largest number of clients. As a result, their AI features are generic, rigid and limited to the data inside their ecosystem. They cannot see across your structured and unstructured data, understand your workflows or tune themselves to your specific GTM motion. And you can’t fine-tune or control the agents they ship.

So you get superficial AI features that summarize data you didn’t need summarized, in formats that don’t serve your teams, with limited ability to customize. If you rely on that as your AI strategy, you are outsourcing intelligence to your SaaS vendors. That’s not a strategy. That’s surrender.

If the solution isn’t tools, what’s the real strategic pivot companies need to make?

You have to flip the model. Instead of buying intelligence, you build the capability to extract it from your own data.

That starts with the data itself. Years before generative AI, Thoughtworks CIO made a bet on the Google ecosystem. In 2023, when Thoughtworks’ marketing couldn't track the health of the brand within traditional SaaS tools, we decided to create a bespoke model, and began onboarding data onto Google BigQuery. The impact of this decision was when AI gained momentum and made its way into the enterprise, it ended up giving us the foundation AI needs: unified, queryable, governed data that agents can reason over. Because this groundwork was in place, rebuilding our core systems was lighter work. We weren’t trying to modernize our data and build agents at the same time.

From there, we pulled business logic out of SaaS tools and exposed it into the Google layer where it could actually be orchestrated by agents. SaaS tools are great for transactions, but they make terrible intelligence systems. You can’t build agents if your logic is trapped in Marketo workflows or Salesforce automations.

Then came talent. We established the AI marketing applications team and introduced forward-deployed AI engineering graduates to work directly with the business. They prototype, fine-tune and iterate on agents daily with marketers sitting right beside them. Some would argue this is shadow IT. 

At the end of 2025, the AI & Data team was elevated from GTM Operations into the office of the CMO, which represents the criticality of this capability in transforming the marketing and sales functions to become AI-powered. The AI & Data team lead by ‘showing not telling’, introducing a culture of live demos, and engaging screen recording using tools like Screen Studio, so that stakeholders stay up to date with progress, rather than having to go through a long list of email updates.

But AI engineers alone aren’t enough. Intelligence without action is just analysis. That’s why we created the GTM engineer role. An evolution of marketing operations focused on workflow automation, system commands, and orchestration in the AI era. They ensure an agent can not only find the signal but also take the action, connecting into systems through n8n and coordinating via Google’s agent-to-agent (A2A) protocol.

The combination of AI engineers for reasoning, GTM engineers for organisational workflow automation is what made the execution of our AI4GTM strategy a reality. This isn’t a tooling strategy. It’s an operating model change.

What does the superagent architecture look like, and why does it matter?

Most companies are creating the agentic version of SaaS sprawl: 50 different agents, each with their own data, logic and UI. It’s the exact mistake we made with tools. We refused to repeat it. We defined our architecture around capabilities, not tasks.

For instance, our superagent, PerformanceAI handles performance intelligence, built on the fact that we already had clean, high-quality performance data. Structured, governed and ready for AI. It was the obvious place to start.

OutreachAI is the superagent that handles prospecting and market engagement, merging internal structured and unstructured data with external intelligence. We invested in our intelligence layers by onboarding tools like people.ai that fill the gap of sales activity, gathering client conversations from sellers inboxes, calendars, meetings and Zoom calls recordings. 

Each capability has clear dataset boundaries, shared business logic and a cluster of specialized agents under a single superagent umbrella.

Technically, this is possible because agents are built using the Google ADK framework, communicate via Google’s A2A protocol and operate over our most critical datasets that are stored in BigQuery. All are supplemented with unstructured data that has been reorganized to work with agents. Making sense of unstructured data was a bigger lift. However our pilot reached 95% accuracy, outperforming enterprise search tools that consistently hallucinated.

And the critical part is the experience layer. We’re moving toward one unified conversational layer that knows:

• Who you are, the persona.

• What work you’re doing.

• Which dataset you can access.

• Which capability you need next.

• And how to trigger actions across systems through workflow automation.

No switching apps, no context loss, no silos. SaaS vendors can’t deliver this because their worldview starts with parts of the story, rather than complete intelligence. While we started our journey with bespoke interfaces for our superagents, we have successfully deployed our first agent onto Gemini Enterprise (GE), which reduces the need for our AI engineers to manage the experience layer, allowing them to spend more time on intelligence and automation. The technology is not where we need it to be, but we are experimenting with a shared experience layer on GE and other technologies, prior to fully migrating out of our bespoke interfaces.

You’ve taken a hard line on “zero UI as the starting point.” Isn’t that risky when most organizations feel safer with dashboards and screens?

Dashboards are a security blanket. They lock intelligence into predefined charts that assume leaders already know what matters. ChatGPT exposed a simple truth: insight doesn’t come from arranging widgets, it comes from asking better questions. Fixed dashboards and fixed buttons share the same flaw. The moment you commit screen real estate, you freeze a mental model and force users to work around it. Our mantra is to keep the screen flexible. Instead of hard-wiring layouts, we fetch formats on demand based on what the user asks. The interface evolves with the question, not the other way around.

Our rule is brutal and simple: strip the interface back until nothing is left but conversation and memory. No dashboards; no widgets; just an agent and a minimal interface that adapts itself to the user.

This approach also accelerates development. If you remove app UI, you remove a lot of complexity, which lets AI engineers focus on the things that matter: data, logic, retrieval, reasoning and actions. Prototyping tools like Bolt and Figma help us 'show, not tell' what the dynamic experience will feel like  without committing to hardcoded UI.

What’s the practical lesson for organizations trying to modernize GTM with AI but stuck in legacy structures?

The hard truth is that if your AI strategy begins with tools, you’ve already lost. The real work starts far deeper. You need a unified foundation where critical structured data lives in one analytical layer and unstructured data is extracted, cleaned and made available for agents. You need to pull your business logic out of SaaS tools and place it somewhere programmable, where agents can actually orchestrate it.

From there, the operating model has to change. AI engineers shouldn’t sit in an IT silo. They need to be embedded in the business, shaping use cases in real time, developing and tuning models from business logic, evolving agents daily. GTM engineers ensure that intelligence becomes action. Position the AI & Data capability as a critical function, linking it closely with business strategy and broader enablement of the marketing organisation. And set a new collaboration model with IT that is suitable for the AI era. 

Instead of creating dozens of disconnected bots, you build super agents centered on core GTM capabilities, each with clear dataset boundaries and shared logic. Those agents need the ability not only to interpret data but to act, reducing SaaS complexity incrementally, starting with more discrete workflows like inbound management and moving to larger systems like CRM.

All of this has to surface through a single conversational experience that removes friction, not adds screens. In a fast-moving environment, you preserve flexibility through one-year vendor contracts, avoiding lock-in and enabling rapid component swaps as the ecosystem evolves.

The organizations that win will treat AI not as an add-on but as an opportunity to rebuild the core of how go-to-market works, where intelligence is unified, actions are automated and value is delivered with human-level quality at dramatically greater speed and a lower cost. That’s how you reimagine value: better outcomes, faster time-to-market, stronger brand reputation and the ability to reinvest productivity gains where it matters.