Whitepaper

    Why AI in sales stalls, and how to build the foundation first

    AI amplifies what is already there: a good system gets stronger, a messy one gets messy faster. This is the order in which to build the foundation.

    • By Sander Vergouwen, co-founder of Sales Surge
    • Updated 16 November 2026
    • 14 minute read

    A typical plan for an AI-Based foundation

    • Phase: Reality checkWeeks: 1 and 2What happens: A scan of data, processes, definitions, follow-up and automationWhat you get: A data quality report, a process map and a gap analysis
    • Phase: Rebuild the foundationWeeks: 3 to 8What happens: Uniform definitions, fixed fields, pipelines with fewer stages and clear criteriaWhat you get: One definition of lead and opportunity, and clear owners
    • Phase: AutomationWeeks: 9 to 14What happens: Workflows for assigning, following up, validating and routing, plus enrichment and reviewed deduplicationWhat you get: A rhythm that does not depend on anyone's discipline
    • Phase: SignalsWeeks: 15 to 18What happens: Website visits, replies and ICP fit translated into priority and timingWhat you get: Priorities based on data rather than gut feel
    • Phase: Human in the loopWeeks: 19 to 22What happens: Deliberately designed moments where a person steps in, with full contextWhat you get: Reps who spend more of their time with customers
    • Phase: Feedback loopsWeeks: From week 23What happens: Dashboards and a fixed review of conversion, signals and routingWhat you get: Small improvements instead of big redesigns

    What is in it

    • Why AI gets stuck

      Dirty data and disconnected tools do not disappear with AI. They only become more visible.

    • AI-Ready or AI-Based

      Putting AI on top of your process, or redesigning the process with AI as the starting point.

    • A plan per phase

      From a first reality check to feedback loops, with a delivery at every phase.

    An AI-Based commercial system as the foundation under B2B growth

    Why do AI projects in sales and marketing stall?

    Because AI amplifies what is already there. A model learns from your data and follows your processes. If those are messy, the AI does messy work, just faster and at greater scale. The technology is rarely the problem. The foundation underneath is.

    Teams often put it like this: we have dozens of tools, dashboards and workflows, and it still feels chaotic. That feeling is accurate. In almost every CRM we open, we find the same four causes.

    Your data cannot be trusted

    Data is a by-product of what people do. Fields stay empty, definitions get mixed up and statuses are set by feel. No automation behaves predictably on data like that, and AI least of all.

    Your processes only exist on paper

    Open the CRM and the process turns out to depend on individuals. Stages get skipped, follow-up varies per rep and intent is not recorded consistently. A process that is not followed the same way every time cannot be automated.

    Your automation is built on exceptions

    Workflows grow out of incidents. A small exception here, an extra rule there, until the whole thing is so brittle that any deviation breaks it. Automation only becomes reliable once you reduce the variation to a set of rules you can actually oversee.

    Your silos pull in different directions

    Marketing chases volume, sales chases conversion and service wants peace and quiet. Each team keeps its own data. Without one shared commercial goal, AI stays a patch on symptoms.

    A familiar scenario. You switch on AI lead scoring in a CRM where a good share of leads has no industry, the stage is wrong and statuses are changed by hand. The model learns patterns that are not real. The score becomes unreliable, the team stops trusting it and the pilot is quietly dropped. That is not bad luck. It is what happens when you build in the wrong order.

    What is the difference between AI-Ready and AI-Based?

    AI-Ready means adding AI to what you already do, like a chatbot or an AI score on top of your current CRM. AI-Based means redesigning the commercial system with automation and AI as the starting point. The first gives you a small step forward. With the second, we design the system so that an extra customer does not mean extra manual work.

    AI-ReadyAI-Based
    Starting pointOptimise what existsRedesign from the foundation up
    Role of AIA bolt-onThe logical result of a stable system
    What you getIsolated pilots and scattered improvementsOne system that improves as a whole
    Role of peopleKeep compensating for what the system missesDo the work where they add value
    RiskPilots that never get past the demoMore upfront work on data and process

    An AI-Based system works because its processes are designed with automation in mind. Where automation does not fit, a person is built in deliberately. Definitions and fields mean the same thing everywhere. Automation sets the rhythm, and AI builds on that rhythm.

    AI-Ready is not wrong, by the way. If you want to see what an agent can do quickly, a scoped AI Sprint is a perfectly good first step, as long as you know the agent runs on the foundations you have today.

    What are the six layers of an AI-Based commercial system?

    Business, data, process, automation, signals and people, each resting on the one below. That is why you start at the bottom and not with the AI.

    1. Business layer: direction drives the design. Your ICP, your segments and goals that connect teams. Without direction, data, process and automation each drift their own way.
    2. Data layer: shared definitions for everyone. One meaning for lead, opportunity and qualified. An owner per field, fixed criteria per stage and minimum standards for completeness.
    3. Process layer: predictability before speed. As few handoffs as possible, clear responsibilities and a CRM that enforces the agreed behaviour.
    4. Automation layer: the work that should run by itself. Assigning, following up, validating, routing, deduplicating and enriching become fixed flows instead of manual chores.
    5. Signal layer: from data to behaviour and intent. Website visits, email replies and ICP fit decide who gets priority and when you call.
    6. Human in the loop: people where they add something. Once the foundation holds, the work shifts to the conversations that matter: the relationship, the nuance, the negotiation.

    The order is not theory. It is the order we build in, because a layer resting on a shaky layer is shaky itself. How the first four layers tie into RevOps is covered on our RevOps implementation page.

    How does an inbound lead move through an AI-Based system?

    Without a shared queue, without anyone googling the company, and without context getting lost on the way. The difference is easiest to see if you follow the same lead twice.

    The same inbound lead, before and after an AI-Based rebuild
    StepBefore the rebuildAfter the rebuild
    ArrivalThe lead waits in a general queueThe record is enriched with industry and size straight away
    AssessmentAn SDR googles the company and guesses the fitAn ICP score on fixed criteria, plus behavioural signals
    AssignmentBy hand, whenever someone has timeAutomatic, by score, segment and capacity
    First contactVaries, sometimes days laterImmediate, with a draft a colleague approves
    HandoffA Slack message to the account managerFull history and a suggested approach in the CRM
    The conversationStarts from scratch, context is gonePicks up where the lead left off

    What happens per layer: the data layer provides fixed fields and minimum data. The process layer removes manual triage. The automation layer handles enrichment, scoring, routing and follow-up. The signal layer sets priority. And the person steps in for the conversation, not the preparation.

    What this does to conversion is not something we will promise up front. It depends on your deal value, your volume and the state of your data. If you get a handful of inbound leads a month, this rebuild will not do much for you yet. We would rather say that now than in month six.

    Where do you start when your data is a mess?

    With definitions, not with cleaning. First agree what a lead, an opportunity and a qualified deal are. Only then do you know which data has to be right and which you can leave alone.

    1. Write the definitions down and have marketing, sales and service sign off. One page is enough.
    2. Set minimum fields per object. What must a new deal contain before it counts?
    3. Clean up duplicates. Surge Dedupe finds them by email, phone and name, and only merges after a review, with an audit trail.
    4. Enrich what is missing from an external source such as Apollo, with fixed rules for which source wins.
    5. Name an owner for data quality. Without one, it will be messy again in three months.

    If you want to know how messy things are today, run the ten checks in the CRM Reality Check. In one morning they show where your data falls shortest.

    How long does it take to build an AI foundation?

    Plan for just over five months to cover every layer, in fixed phases with a delivery at the end of each. After week 8 the definitions, fields and pipelines are in place. The first workflows run around week 14. This is how we usually plan it. Your timeline depends on the number of teams, pipelines and data sources.

    We invoice per phase, before that phase starts. So after every delivery you decide whether to continue.

    A view ahead: what does commercial work look like towards 2030?

    This is a point of view, not a measurement. It is how we expect the work to shift as systems become more consistent. We see early signs in our projects, but we are not going to put percentages on it.

    • The line between marketing and sales blurs, because both work from the same data and definitions.
    • AI takes over the preparation: intake, research, routing, scoring and the first follow-up.
    • Outbound starts from buying signals rather than from a list.
    • A first draft of an analysis or proposal comes from AI. People add the interpretation and the conversation.
    • Human work concentrates on the moments that require trust.

    That shift only happens if the foundation is already in place.

    Who is this plan not for?

    If you have fewer than five salespeople, start with a clean pipeline, not with this plan. A well-configured CRM with a few fixed workflows will do more for you than six layers of design.

    It does fit B2B teams that want to get past the pilot stage and grow revenue with the team they already have. Not sure where you stand? Pick one of these:

    Review of a CRM setup

    No more follow-ups from memory

    “I spend my days fixing conversion leaks for ecommerce brands, so I'm picky about who touches my own systems. Sales Surge passed the test.”

    Ildebrando Mancini

    Convert Design, Oct 2026

    Read the review on the Pipedrive Marketplace
    Pipedrive Platinum Partner

    Pipedrive Platinum Partner

    View partner profile

    In short

    An AI-Based commercial system is a sales and marketing operation designed from the ground up with automation and AI as the starting point, rather than AI bolted onto an existing process. The core idea is simple: AI amplifies what is already there. A good system gets stronger, a messy one gets messy faster. Below is the order in which to build it.

    Frequently asked questions about the AI-Based foundation

    What is an AI-Based commercial system?

    A sales and marketing operation designed from the ground up with automation and AI as the starting point. Instead of adding AI on top of your current process, you redraw your data, workflows, signals and the work your people do, so AI simply runs as part of it. We design it so that an extra customer does not mean extra manual work.

    What is the difference between AI-Ready and AI-Based?

    AI-Ready means adding AI to what you already do, such as a chatbot or an AI lead score on top of your current CRM. AI-Based means redesigning the whole commercial system with AI and automation as the starting point. AI-Ready gets you a small step forward. AI-Based asks for more upfront work on data and process, and gives you one system that improves as a whole.

    Why do most AI initiatives in sales and marketing stall?

    Because AI amplifies what is already there. Dirty CRM data, stages nobody defines the same way and tools that do not talk to each other do not go away with AI. They just become more visible. An AI lead score built on bad data is still a bad score. So foundations first, then AI.

    Who is this approach for?

    B2B teams that want to get past the pilot stage and grow revenue with the team they already have, in SaaS, professional services, manufacturing or agencies. With fewer than five salespeople it is too early. Start with a clean pipeline and a properly configured CRM instead, and we will tell you exactly that on the intro call.

    How long does it take to build an AI-Based foundation?

    After the first two phases, around week 8, the definitions, fields and pipelines are in place. The first workflows run around week 14, at the end of the automation phase. All six layers take until roughly week 22, after which the feedback loops continue as ongoing work. We deliver and invoice per phase, so you do not have to wait until the end to see whether it works.

    What role does Sales Surge play?

    We are a Pipedrive Platinum Partner, we also implement HubSpot and Zoho, and we build the foundation together with your team. In practice: the data and CRM layer, the workflows and automation, AI agents for research, scoring and outreach, and the RevOps agreements around them. Want it fully outsourced? We run it. Prefer to do it yourself? We deliver the blueprint.

    Start with the reality check in weeks 1 and 2

    We map your data, processes and automation, and you get a data quality report, a process map and a gap analysis. Book a call and we will look at where your foundations stand.