All the Gear, No Idea: What Makes AI Tools Work in GTM?
- 3 hours ago
- 5 min read
In our previous article, we looked at where AI is already adding value across go-to-market teams, from content creation and lead qualification to sales call analysis and customer feedback.
But buying an AI tool is the easy bit. Making it work is much harder.
Across the implementation of AI we have seen in marketing, sales and customer teams, the difference between success and disappointment rarely comes down to the tool alone. More often, it comes down to the use case, the quality of the data, the way the tool is implemented, and whether the team actually change how they work.
From our work with clients, we have seen six themes that are strong predictors of whether an implementation of AI is likely to create value.
Implementation beats technology
This is the biggest lesson: whether a business chooses an in-house build or a third-party tool, success depends on whether the team actually uses it in a way that improves performance.
There is a useful parallel with the early days of workplace computing. By the late 1980s, companies had invested heavily in computers, but the promised productivity gains weren't materialising. The issue was not that computers were useless: it was that many businesses had added them to existing ways of working, rather than redesigning workflows around what the technology made possible.
Spreadsheets are a good example. Before tools like VisiCalc, Lotus 1-2-3 and later Excel, many business calculations were done through paper ledgers, desktop calculators and manually updated schedules. The breakthrough was not just faster arithmetic. It was that formulas allowed a whole model to recalculate when an assumption changed. The real productivity gains came when teams stopped treating spreadsheets like electronic paper and started redesigning planning, reporting and analysis around dynamic models.
This became known as the productivity paradox, captured by Nobel laureate Robert Solow’s famous observation that the computer age was visible everywhere except in the productivity statistics. We expect AI adoption to follow a similar pattern.
The companies that get value from AI will not always be the ones with the most cutting-edge tools. They will often be the ones with clear ownership, good data, sensible workflows, training, incentives and a way to measure whether adoption is happening in practice.
A slightly less sophisticated tool that is embedded into a team’s day-to-day work will beat a more advanced tool that nobody uses properly.
This is where Revenue Operations becomes increasingly important. RevOps teams are often best placed to connect the tool, the data, the process, human behaviour and the commercial outcome.
Scale matters
AI in GTM is most effective in automating simple, well-understood and repeatable tasks, as with many other functions. The less standardised the process, the harder it becomes for AI to deliver reliable results.
For example, AI can be useful for reading across hundreds of sales-call transcripts and categorising customer sentiment, objections or competitor mentions. That is a repetitive task where the model can help process more information than a human team could review manually. The task itself can scale, making it the easier to justify the upfront investment and realise meaningful commercial value.
Where it becomes harder is when the task moves from classification to judgement. Deciding whether a lead should be followed up, how it should be prioritised, or what the next best action should be often requires more context than the transcript alone can provide.
AI tools have not yet consistently demonstrated that they can interpret numbers, tone, nuance, timing and commercial context as reliably as experienced professionals.
AI can help surface useful signals, but for now, it should not always be asked to make the decision.
AI is better with qualitative data than quantitative data
Some of the strongest AI use cases we see involve unstructured qualitative data: sales calls, emails, support tickets, reviews, RFPs, survey responses and prospect research.
This makes sense: Large language models (LLMs) are designed to work with language. They are good at summarising, classifying, rewriting and finding patterns in text.
They are less generally reliable when asked to do statistical analysis, prove causal relationships or conduct hypothesis testing in quantitative datasets. Asking an LLM to determine which lead attributes statistically predict conversion is a very different task from asking it to summarise the most common objections in sales calls.
That does not mean AI has no role in quantitative analysis. But for most GTM teams today, the clearer value is in helping structure messy qualitative information that was previously difficult to summarise systematically or to analyse consistently.
We have not seen strong evidence of AI outreach working
There is a lot of noise around AI-generated outbound, AI SDRs and fully automated sales outreach. So far, we have not seen much evidence of this working well among our clients.
To be clear, algorithmic outreach existed long before generative AI. What AI has added, in many cases, is the ability to produce longer, more personalised-sounding emails at much greater volume.
That is not always a good thing.
Under pressure to hit activity targets, outbound teams can easily end up sending large volumes of AI-generated emails that are verbose, generic or overly familiar. The result is often more spam, not more pipeline.
AI can support parts of the sales process, but replacing thoughtful, relevant outreach with mass-generated copy is unlikely to improve the buyer experience – certainly in the short term.
Automation can be valuable, but it is not the same as AI
Many of the most effective improvements we see in GTM teams are not really AI at all: they are automation.
Rules-based workflows, lead routing, triggered emails, CRM updates and process automation can all create significant value. They are often easier to govern, easier to explain and easier to verify than AI-generated outputs.
That distinction matters. If a workflow is transparent, repeatable and the desired output is clear, a rules-based automation may be the better solution.
The question should not be, “How do we use AI here?” It should be, “What is the best way to improve this workflow?” Sometimes the answer will be AI. Often, it will be automation. Occasionally, it will be fixing the underlying process first.
AI adoption needs the right culture
Finally, successful AI adoption needs to come from the top, but not as a vague instruction to “do more with AI”.
A monthly private equity board question about where AI is being adopted is not the same as a strategy.
Teams need to understand why AI is being introduced, what problem it is meant to solve, and how it will help them do better work. The people expected to use the tools need to be involved in the discussion, not simply handed a login and told to get on with it.
Without cultural support, clear ownership and a compelling reason to change behaviour, even promising AI tools can end up with limited adoption and fragmented impact.
The bottom line
AI will continue to reshape go-to-market teams, but value will not come from adoption for adoption’s sake.
For private equity investors and portfolio companies, the priority should be to identify where AI can genuinely contribute to value creation: improving speed, consistency, insight or conversion in specific parts of the customer journey.
The businesses that benefit most will be those that start with a clear commercial problem, choose the right tool for the job, implement it properly, and measure whether it is actually being used.
If you would like to discuss how to choose the right tools, implement them effectively, and measure AI adoption across your GTM function, please get in touch.
All views expressed in this post are the author’s own and should not be relied upon for any reason.
