Diogo Silva

The end of the information broker

Artificial intelligence will not replace the person who shares the risk of a difficult decision. It will replace the salesperson who spends the week managing administrative noise instead of sitting across the table from the client.

When I wrote on LinkedIn recently about the future of the B2B Account Executive over the next five years, my aim was not to add to the usual excitement surrounding artificial intelligence. It was to point out a practical contradiction that became clear to me while building my own internal tools: in complex enterprise sales, the highest use of technology is not to replace human contact, but to buy back the time required for it.

The prevailing consensus in the software industry assumes that artificial intelligence exists to increase volume. Every week brings a new promise of autonomous prospecting agents, automated cold email sequences, and synthetic personalisation at scale. The result is familiar to anyone with a corporate inbox: thousands of machine-generated messages sent by one algorithm, only to be filtered out seconds later by another algorithm on the buyer’s side. We have automated the production of noise.

Complex enterprise decisions have never worked that way, and they are not about to start.

No Chief Financial Officer or HR Director restructures the compensation, insurance, or tax-exempt benefits of five hundred or a thousand employees because an automated sequence caught them at the right moment. Decisions of that scale alter the monthly net income of real families, carry strict fiscal and regulatory scrutiny, and involve delicate internal politics. When an executive signs a six-figure contract, they are not merely buying software; they are taking a personal and institutional risk. Nobody takes that step without looking someone in the eye who understands their operation, knows the law, and is willing to share the responsibility for the outcome.

The real threat to the enterprise salesperson is not that a machine will close the deal in their place. The threat is operational asymmetry: continuing to spend half the working week on manual preparation and administrative upkeep while a systems-native peer manages the same portfolio with twice the analytical depth and three times the presence in the field.

1. The death of the information broker

For twenty years, a large part of software sales consisted of acting as an intermediary between the product and the buyer. A prospect requested a meeting to find out what the platform did, how much it cost, and how other companies used it.

That role is finished. By the time a corporate buyer enters a first exploratory call today, they have already completed roughly eighty per cent of their research independently. If a salesperson uses a forty-five-minute meeting to walk through a standard catalogue demonstration or recite features that are already public, they are competing directly with a self-serve webpage—and losing.

To earn a seat at the table now, you must arrive with a diagnostic of the client’s financial and operational reality that is sharper than their own internal synthesis. A CFO does not need a tour of a user interface; they need to see a rigorous model of their current tax inefficiencies, the exact return on investment, and a clear path through their internal bottlenecks. Either you bring genuine commercial insight into the room, or you are merely a delay in their calendar.

2. The film and the snapshot: a three-layer architecture

Reaching that level of preparation across dozens of active enterprise accounts usually creates a severe bottleneck. Context is scattered across old CRM notes, email threads, annual financial reports, hiring signals, and statutory tax rules. Spending forty-five minutes jumping between browser tabs before every call is unsustainable; skipping the preparation altogether is amateurish.

Over the past few weeks, I designed and deployed an internal Deal Intelligence engine to solve this problem in my own workflow. Building a system for daily production teaches you very quickly what is theatre and what actually works. The architecture rests on three operational layers.

Layer 01: A strict hierarchy of sources. Search engines give you a static snapshot of a company; the CRM holds the motion picture. Public web data can tell you a company’s declared revenue, its corporate group structure, or how many people it lists on LinkedIn. Useful as that is, deals are won or lost inside the historical film of the account: who the incumbent provider is, which insurance broker holds political influence, why a conversation stalled fourteen months ago, and who actually holds the veto. My system enforces a strict reading order:

  1. Historical CRM meeting notes (incumbents, political blockers, and real decision-makers).
  2. CRM email threads (proposals sent, unspoken objections, and patterns of silence).
  3. Public web and hiring signals (corporate group structure, leadership changes, and HR expansion).
  4. Internal statutory and product wiki (tax legislation and platform boundaries).

Layer 02: AI reads and connects; deterministic code does the maths. One of the most dangerous mistakes in commercial operations is asking a large language model to calculate financial returns or tax savings. Generative models are probabilistic; tax law is not. When you are modelling Portuguese Social Security exemptions (TSU at 23.75%), statutory meal allowance thresholds (€6.00 in cash versus €10.20 per day on a meal card), or corporate tax deductions, a hallucinated figure destroys executive trust instantly.

In a sound architecture, the language model is used exclusively for reading dense unstructured history and auditing qualification criteria—evaluating all eight pillars of MEDDPICC (Validated, Partial, or Missing) across the entire portfolio every week and isolating the weakest link in each deal. Every single financial calculation, alongside CRM data-hygiene checks (flagging mismatched subsidiary tax IDs, duplicate opportunities, or stale headcount figures), is handed over to fixed, deterministic code. Zero error is the only acceptable standard when presenting a business case to a finance director.

Layer 03: The thirty-second pre-call brief. The output of a system should never be another long report that nobody has time to read. Before stepping into a meeting, the engine surfaces a structured brief that can be absorbed in thirty seconds: a one-sentence diagnostic at the top of the account, the primary political or operational blocker, a single concrete next step, and two distinct executive angles—a financial efficiency case for the CFO and a talent-retention, low-friction case for the HR Director.

3. The golden rule of pipeline: Temperature ≠ Account Size

The second major failure in traditional sales management is confusing the size of an account with the health of the opportunity. A stalled hundred-thousand-euro deal is not a “hot” opportunity; it is simply a large distraction that flatters a spreadsheet while starving the forecast of reality.

To remove wishful thinking from pipeline reviews—the familiar trap of forecasting a deal because “the call went really well”—the system separates Tier (potential annual contract value) from Temperature (objective velocity and qualification, scored from 0 to 10):

  • Validated MEDDPICC (up to +5.0 points): Each validated pillar adds weight, but Champion and Economic Buyer carry double weight (2x). In enterprise cycles, having an internal advocate with personal capital at stake increases conversion more than threefold.
  • Pipeline velocity (up to +2.0 points): Measured against real historical benchmarks. In our enterprise segment (companies with 500+ employees), the median cycle from qualified opportunity to signature is roughly 74 to 85 days, with half of all winning deals closing between 48 and 148 days. Deals moving within this cadence earn velocity points; deals dragging past the upper quartile lose them.
  • Recent sign of life (+1.5 points): Verified two-way communication—an inbound client email or completed meeting—within the last 14 days.
  • Firm next step (+1.5 points): A confirmed meeting on the calendar within the next 14 days.

Crucially, the engine enforces an automatic Safety Brake: if an opportunity has no validated Champion, or if the client has not replied in more than 25 to 30 days, the deal is mathematically capped at Warm (below 7.0), regardless of how prestigious the logo is or how much revenue is at stake. It forces honesty before an opportunity quietly dies of neglect.

4. When the digital is commoditised, physical presence becomes a luxury

Why go to the trouble of engineering this level of automation?

If the goal were simply to sit behind a monitor and manage more video calls per day, it would miss the point entirely. I have written before about how efficiency can become an end in itself, and about the importance of retaining the ability to leave work when the day is done. A tool should serve the craft, not consume the worker.

When every competitor can generate endless digital noise and schedule back-to-back fifteen-minute video calls, the digital channel loses its power to differentiate. Scarcity shifts to the physical world.

Cutting pre-call research and CRM administration from forty-five minutes to thirty seconds buys back roughly sixty per cent of the administrative week. That reclaimed time is what allows me to close the laptop, get into the car, drive to an industrial estate or a company headquarters, and sit in a room with a Finance Director and an HR team. It creates the unhurried space required to listen properly, read the tension in the room, and help multiple decision-makers align around a complex change.

The future of enterprise sales does not ask us to choose between the traditional relationship seller and the systems engineer. The rule for the years ahead is simpler and more demanding: automate everything that happens before and after the meeting so that you can be entirely human—and irreplaceable—during the meeting. You become the engineer of your own system precisely so you can afford the time to be the best salesperson in the field.

Reply by email

Letters to the editor

Write a letter

Your email stays private. Letters are reviewed before publication.