Agentic Commerce B2B: closing 5 data gaps in 2026

The figures are unambiguous: within a single year the use of autonomous AI agents among mid-sized B2B companies has almost doubled – from 8.5% to 16.6%. According to the current study “Agentic Readiness im B2B Commerce” by dotSource and ECC Köln, a further 37% of companies plan to introduce or expand AI agents in 2026. Enthusiasm is high and the technology is available. Yet one problem remains: the data foundation is missing.

88% of the companies surveyed see considerable ground to make up on the seamless integration of ERP and CRM systems. 83% criticise the lack of standardisation of product data and prices. 82% struggle with fundamental problems of data quality. Agentic commerce – the use of autonomous AI agents for purchasing, ordering and sales management – only works if the data foundation is sound. Otherwise AI agents remain toys without effect.

This article shows mid-sized manufacturers and wholesalers which five concrete data gaps they have to close so that AI agents – their own as well as those of their customers – can work reliably. Each gap comes with one clear practical step.

What is agentic commerce, and why now?

Agentic commerce describes the use of autonomous software agents that take on tasks in e-commerce independently. In a B2B context this means AI agents analyse order histories, check stock levels, compare prices, negotiate terms and place orders – without a person having to steer every step manually.

The concept is not new. Automated ordering systems and inventory management interfaces have existed for years. What is new is the quality of the autonomy: modern AI agents learn from data, adapt to context and take decisions that used to be taken by buyers or sales staff.

For manufacturers and wholesalers two scenarios arise. Internal agents automate sales processes, pricing, reordering and customer analysis. External agents, by contrast, are deployed by customers to buy from suppliers autonomously – anyone who is not “agent-ready” is no longer found. Where AI agents deliver real ROI in sales and where caution is called for is explored in more depth in the article on AI agents in B2B sales.

The pressure is rising. Anyone not prepared for agentic commerce in 2026 loses market share to competitors that deliver faster, calculate more precisely and are easier to find.

Data gap 1: missing ERP and CRM integration

The biggest hurdle for agentic commerce is the missing seamless connection between ERP, CRM and shop systems. 88% of companies state that their systems are not integrated end to end. That means prices, stock levels, customer data and order histories sit in different systems – often without real-time synchronisation.

An AI agent can only work as well as the data it has access to. If stock levels are managed in the ERP, prices in the shop and customer histories in the CRM, the basis for autonomous decisions is missing. The agent cannot check whether a product is available, which price applies to a particular customer or which payment terms have been agreed.

Practical step: think API-first. Rather than maintaining isolated solutions, companies should rely on open interfaces that connect ERP, CRM, PIM (Product Information Management) and shop systems. In concrete terms this means:

Real-time synchronisation of stock levels, prices and customer data

Central data management for product information (PIM as the single source of truth)

Middleware solutions that orchestrate the different systems

Many modern ERP systems offer REST APIs that make integration easier. Which hurdles typically arise when connecting ERP, PIM and shop is set out in the article on B2B system integration.

Data gap 2: inconsistent product data and prices

83% of companies see ground to make up on the standardisation of product data and prices. The problem: product names vary from channel to channel, prices are maintained manually, attributes are missing or incomplete.

An example: a manufacturer of industrial components lists the same product in three systems under different article numbers. In the ERP it is called “screw M8x40”, in the shop “hexagon screw 8mm”, in the CRM “SKU-12345”. For a person that is still followable. For an AI agent it is data chaos. On top of that, prices often vary by customer, quantity, region or sales channel. Without clear rules and data structures, an agent cannot produce reliable quotations or place orders.

Practical step: introduce a PIM and define data standards. A product information management system serves as the central data foundation. All product information is maintained there – consistent, complete and structured. Concrete measures:

Unique article numbers (GTIN, EAN or internal SKU) for every product

Complete attribution: dimensions, weight, material, colour, technical data

Price rules instead of price lists: define how prices are calculated (e.g. volume discounts, customer segments, regional surcharges)

Classification according to standards such as eCl@ss or ETIM for technical products

A PIM not only reduces sources of error, it also speeds up time to market for new products. When the introduction really pays off for mid-sized companies is answered in the article on introducing Akeneo PIM.

Data gap 3: poor general data quality

82% of companies struggle with fundamental problems of data quality: missing data, outdated information, duplicates, inconsistent formats.

A classic example: a customer is held in the CRM three times over – once as “Müller GmbH”, once as “Müller & Co.” and once as “H. Müller Handelsgesellschaft”. Each record has different addresses, contacts and payment terms. For an AI agent it is unclear which record is the right one. It is much the same with product data: missing images, incomplete descriptions, outdated technical data sheets. An agent cannot sell what it cannot describe.

Practical step: establish data governance. Clear rules determine who maintains data, how it is structured and who is responsible for quality. Concrete measures:

Data audit: take stock of all data sources and quality problems

Define responsibilities: who maintains product data? Who updates customer data?

Automated validation: mandatory fields, format checks, duplicate detection

Regular clean-up: remove old, inactive or faulty records

Data governance is not a one-off project but a continuous process. Investing here creates the basis for customer analysis, personalisation and sales transformation.

Data gap 4: missing first-party data and customer insights

Many B2B companies know surprisingly little about their customers. Order histories sit in the ERP, enquiries in the CRM, website visits in web analytics – but nowhere is this data brought together and analysed.

The problem: without first-party data and customer insights, AI agents cannot produce personalised offers. They do not know which products a customer orders regularly, which prices that customer accepts or which payment terms they prefer. An example: a wholesaler of electrical engineering products has supplied a customer for ten years. The customer orders the same components every three months. An AI agent could prepare this order automatically, propose the best delivery date and offer a volume discount – if the data were available and connected.

Practical step: build a customer data platform (CDP). A CDP brings together all customer data from ERP, CRM, shop and marketing. Concrete measures:

Unified customer profile: all data on a customer held in one profile

Behavioural analysis: which products are bought? Which pages visited? Which enquiries made?

Segmentation: cluster customers by buying behaviour, revenue potential and industry

Predictive analytics: forecast when a customer will reorder or which products might be of interest

A CDP is the basis for personalisation and omnichannel sales. AI agents draw on this data and take decisions that rest on real customer behaviour – not on assumptions.

Data gap 5: unclear pricing models and a lack of transparency

Pricing in B2B is complex: volume discounts, customer segments, regional surcharges, special terms, campaign prices. Many companies maintain these prices manually in Excel lists or negotiate them individually with every customer.

For an AI agent that is a problem. It cannot calculate prices autonomously if the rules are not clearly defined and digitally mapped. The result: the agent has to involve a person for every enquiry – and with that the autonomy is lost. A further problem is the lack of transparency. When prices cannot be followed, customers lose trust.

Practical step: digitalise and automate pricing models. Concrete measures:

Rule-based pricing: define price rules (e.g. “from 100 units: 10% discount”)

Prices by customer segment: different price lists for different customer groups

Dynamic price adjustment: adjust prices on the basis of stock levels, demand or competition

Create transparency: show customers how prices come about (e.g. “volume discount: -10%”)

Modern ERP and shop systems offer functions for rule-based pricing. How customer-specific prices can be managed without chaos is described in the article on pricing governance for B2B shops.

Agentic commerce as a “human in the loop” scenario

One important point: agentic commerce does not mean that people become superfluous. Quite the opposite – the most successful models rely on “human in the loop”: AI agents take on routine tasks, people take the strategic decisions.

An example: an AI agent analyses order histories, identifies a customer who reorders regularly and prepares an order. A sales employee checks the quotation, adjusts it where necessary and releases it. The agent saves time, the person safeguards quality. This model only works if the data foundation is sound: if ERP, CRM and shop are integrated, product data is standardised, prices are transparent and first-party data is available.

Why act now?

The study shows that 37% of companies plan to introduce or expand AI agents in 2026. Anyone who fails to act now falls behind. Customers increasingly expect orders to be handled quickly, precisely and automatically. Competitors that rely on agentic commerce deliver faster, calculate more accurately and are easier to find.

On top of that, the technology is available. AI agents are no longer science fiction but reality. The question is not whether agentic commerce is coming, but when a company will be ready for it. Closing the five data gaps is not a sprint but a marathon. Every step, though, brings manufacturers and wholesalers closer to a future in which sales is more efficient, more scalable and driven by data.

First steps: from analysis to implementation

Many companies know they have to act – but not where to start. One proven approach:

Take stock: which systems are in use? Where does the data sit? Which interfaces exist?

Set priorities: which data gap has the greatest influence on revenue and efficiency?

Identify quick wins: which measures can be implemented quickly and deliver value straight away?

Develop a roadmap: which steps follow in the medium and long term?

Bring in partners: which expertise is missing internally? Who can provide support?

Getting started with agentic commerce does not have to begin with a million-euro project. Often it is enough to start with one integration, introduce a PIM or digitalise price rules. What matters is that a start is made.

Conclusion: data quality decides success

Agentic commerce is no longer a vision of the future but reality. AI agents already take on tasks that buyers and sales staff used to handle. Without clean data, however, they remain ineffective.

The five data gaps – missing ERP and CRM integration, inconsistent product data, poor data quality, missing first-party data and unclear pricing models – are the biggest hurdles. Closing them creates the basis for data-driven sales models, personalisation and sales transformation. The technology is available. The question is: is your data ready?