← All insights
AI & Commercial Transformation · 5 min read
AI in the commercial function: what we have learned, and what usually goes wrong
The question most boards are asking
Every PE-backed business I speak to is asking some version of the same question: which AI tools should we be using in the commercial function? It is a reasonable impulse. The technology is moving fast, the pressure to modernise is real, and the fear of falling behind competitors who are deploying it is genuine. But it is the wrong starting point, and answering it first tends to produce a predictable outcome: a set of expensive subscriptions, low adoption, and a leadership team that has not moved any commercial KPI.
The better question is this: where in your commercial process does poor-quality information cause the most expensive decisions? Answer that first. Then AI has a precise job to do. Without that answer, you are introducing a powerful accelerant into a system you do not fully understand.
The businesses that get the most from AI in commercial functions are not the ones that deploy the most tools. They are the ones that are most precise about the problem they are solving before they introduce any technology.
Where we always start
Before any AI tool enters a Hirondl engagement, we run a data diagnostic. Three questions: What commercial decisions is this business making every week? What information are those decisions based on? And how reliable is that information?
In most cases, the answer is uncomfortable. Transaction data is incomplete. CRM adoption is patchy. Pricing decisions are made by individual reps using instinct rather than data. Customer segmentation reflects history, not potential. AI introduced into that environment will not fix it. It will amplify it, and faster than a team of analysts would have.
Where AI creates value in the commercial function
01
Pricing diagnostics
Most B2B businesses do not know which customers are profitable at the transaction level. AI can scan years of invoice data to surface structural discount patterns, price-cost gaps by customer, and margin variation by product group that manual analysis would take months to find, and that rep intuition consistently misses. The output is a ranked list of recoverable margin by account and product group, ready to act on within the first sixty days.
→ Pricing Transformation case study
02
Customer profitability by account
Profitability calculated at the invoice level routinely overstates margin. An account that looks healthy on revenue can be destroying value once freight, credit terms, customisation, and support costs are allocated against it. AI-assisted cost-to-serve modelling builds a true profitability view across the entire customer portfolio, segmented by the levers that are actually controllable. That changes the conversation in every commercial review and every pricing negotiation.
→ Pricing Transformation case study
03
Prospecting and territory design
AI-assisted ICP scoring and firmographic matching can build a qualified prospect pipeline at a scale no sales team could sustain manually. Combined with territory design, matching rep capacity to addressable opportunity, the result is a prospecting infrastructure the commercial team can run continuously, not a one-off list that goes stale. The insight that a business is sitting on 46,000 tonnes of untapped opportunity is only actionable if you can identify who to call.
→ Specialty Films case study
04
Pipeline health and sales performance
Pattern recognition across historical deal data produces scoring that flags at-risk opportunities before they slip and forecasts that are data-driven rather than rep-reported. At the rep level, AI-generated coaching summaries surface the specific behaviours, objection handling, follow-up cadence, deal velocity, that separate high performers from the rest. The insight already exists in the CRM and call data most commercial teams are sitting on. The question is whether you are extracting it.
05
Market intelligence and competitive positioning
Synthesising competitor pricing signals, market share proxies, and product positioning across multiple geographies is a task that previously required months of desk research. AI compresses the synthesis phase to days: regulatory summaries, competitive intensity proxies, and distributor landscape pre-screening, all structured and ready for strategic interpretation. The decisions remain human. The time to reach them is cut significantly, which, inside a PE hold period, is often the constraint that matters most.
→ Asian GTM Strategy case study
2–4pp
gross margin improvement typical from AI-assisted pricing diagnostics in the first twelve months
60 days
from data to ranked customer profitability view, the foundation of every commercial transformation
400+
qualified prospects built in one engagement using AI-assisted ICP scoring and firmographic matching
What bad implementation looks like
AI introduced into a broken commercial process does not fix it. It amplifies it, and faster than a team of analysts would have.
We have seen conversational intelligence platforms deployed before anyone had defined what a good sales call looks like. Prospecting AI tools activated before the business knew which customers were actually profitable. Outbound copy generated at scale before a validated message existed.
In each case, the result was the same. Volume went up. Quality went down. The commercial function became harder to manage, not easier, because AI had removed friction from an activity that was already pointed in the wrong direction. The failure mode is always predictable: treating AI as a solution to a problem you have not yet defined.
There is also a softer failure mode, harder to see in the moment. When a commercial team adopts AI tools as a productivity layer before the underlying commercial process is sound, they often lose the discipline of doing the diagnostic work manually. They skip the step of understanding the data because the AI generates an output quickly. The speed creates confidence. The confidence is often misplaced.
The PE timing constraint changes everything
Most AI implementation advice assumes you have eighteen months to run pilots, iterate, and find what works. PE-backed businesses rarely do. When you have a thirty-six-month hold period and EBITDA improvement is expected in year one, the sequencing of AI deployment is not a technology question, it is a programme management question.
In our experience, AI delivers the highest return in the diagnostic and data structuring phase, the first sixty days of an engagement. This is where slow analysis is most expensive, where speed has the most direct commercial consequence, and where AI has the least risk of amplifying a broken process. Because at that stage, the process is being rebuilt from the data up.
Later in an engagement, AI supports sustainment: coaching summaries that update automatically before manager sessions, pipeline dashboards that flag outliers in real time, prospecting engines that run continuously rather than in quarterly bursts. But the strategic diagnosis has to come first. And the strategic diagnosis has to be human.
How we sequence it
01
Diagnose, identify where poor-quality information causes the most expensive commercial decisions
02
Build the foundation, clean CRM data, consistent lifecycle definitions, baseline commercial metrics
03
Deploy precisely, apply AI specifically to the identified gap, not as a general productivity layer
04
Embed into workflow, build AI outputs directly into the daily process so adoption is automatic, not effortful
05
Validate before expanding, confirm commercial KPIs are moving before broadening the AI scope
The short version
AI is genuinely useful in commercial transformation. We use it in every engagement we run. The pattern it follows in good implementations is consistent: it compresses the time between raw data and actionable insight, it creates tools the commercial team continues to use after we leave, and it lets our partners spend their hours on judgement rather than on information gathering.
The pattern in poor implementations is equally consistent: a tool was introduced before the problem was defined. The question you need to answer first is not which AI tools to buy. It is where in your commercial process you are flying blind, and what it costs you every quarter that you are.
Working through an AI deployment challenge in a portfolio company?