9 best practices for enterprises implementing and scaling agentic commerce

Introduction

Once considered a futuristic concept born out of sci-fi imaginations, agentic commerce — where autonomous AI agents transact, recommend and act on behalf of consumers — is rapidly becoming a reality. 

Consumers are already getting comfortable with this new paradigm. McKinsey reports that 50% of consumers already use AI-powered search today, and estimates that $750 billion of consumer spend will flow through AI-powered search by 2028.  

Forward-thinking retailers like Frasers Group and Liverpool are already preparing for this shift, meeting customers where they increasingly spend time: Within GenAI-powered channels such as ChatGPT, Copilot and Gemini.

But while the opportunity is massive, so is the risk of rushing in without a strategy. Implementing agentic commerce is an evolution that touches data architecture, governance, human workflows and customer experience, so it requires more than a quick-win mindset. 

Here’s a comprehensive guide to the best practices for enterprises looking to implement and scale agentic commerce effectively.

Best practice #1: Start with your data foundation

You can’t expect an AI agent to make intelligent, brand-safe decisions if it’s drawing from inconsistent or incomplete data. Just like humans, agents can only act on information they can trust.

Building a clean, connected and machine-readable data foundation is the essential first step for any enterprise entering agentic commerce. This means:

• Centralizing core data (product, pricing, inventory, and policy) across all commerce systems.
• Standardizing formats to make information structured and machine-readable. Product details should be expressed in structured schemas and rich attributes, rather than being hidden in PDFs or unstructured copy, in order to ensure discoverability. Standardizing promotions and offers also enables AI agents to easily parse, compare and act on them autonomously.
• Exposing open APIs that enable agents to access live, trustworthy data on availability, delivery promises and pricing.
• Audit and structure first-party data to capture preferences, usage patterns and post-purchase behavior in formats optimized for agent consumption.
• Implement event-driven architectures that enable agents to trigger automated workflows, such as restock alerts, reorders or service updates, in real-time.
• Create privacy-safe data pipelines that share user preferences and feedback with agents while maintaining full regulatory compliance.
• Conduct agent-readiness assessments to evaluate how your platform performs under simulated agent interactions and workflows.

For example, before your products can be discovered or purchased through ChatGPT or any GenAI channel, you must ensure that your product catalog is accurate, your pricing reflects real-time availability and your inventory systems can provide instant confirmation. 

Ask yourself: What data would you need to expose in the next six months to make your offers complete, trustworthy and discoverable by agents?

And finally, even if you start small — say, by deploying agents for a single category or region — it’s vital to build a scalable data foundation. Clean, structured data doesn’t just support your first agentic use case; it unlocks the unexpected value of intelligent automation across the enterprise.

Best practice #2: Empower your people in a hybrid human-AI environment

According to McKinsey, “effective and scaled agent deployments could deliver productivity improvements of three to five percent annually and potentially lift growth by 10 percent or more.”

This is huge — but it can’t be accomplished with AI agents alone. For businesses to manage this increased output in productivity, they need to find a balance between humans and machines, and the roles of each. Enterprises that thrive in the agentic era will be the ones that treat AI not as a replacement for people, but as an enabler of their expertise and creativity. 

While it’s expected that AI agents take on the repetitive, data-intensive tasks, human oversight and governance are needed to avoid “agent chaos” that may arise, especially in large-scale organizations operating across multiple business units, brands or regions. That makes the collaboration between humans and agents even more critical. 

Here’s what that looks like in practice:

• Upskill for collaboration, not competition. Equip teams with the skills to understand, monitor and improve AI systems. Invest in agentic fluency across the organization as a business capability.
• Give employees agency over AI. Create “AI champion” roles within business units to experiment with new use cases, document learnings and share success stories.
• Reward experimentation. Encourage teams to test, learn and iterate. Agentic commerce evolves rapidly, and organizations that reward curiosity rather than punishing failure learn more quickly than their competitors.
• Tap into feedback. People on the front lines, such as store managers to customer service representatives, often see where agents fall short. Build mechanisms for them to flag issues and co-create improvements with your AI teams.

When people feel empowered, AI becomes a force multiplier rather than a source of disruption. The result is a smarter, more agile organization where humans focus on what only humans can do — creativity, judgment, connection — while AI scales execution and insight.

Best practice #3: Begin with strategy, not technology

Before deploying agents, step back and define what success actually looks like for your business. Too many enterprises start with a pilot that’s technologically impressive but strategically irrelevant.

A strong agentic strategy answers key questions such as:

• What customer or business problem are we solving with agentic capabilities?
• How will agents create measurable value: Revenue growth, efficiency or customer retention?
• What roles will humans continue to play alongside agents?
• How do we ensure our brand remains visible and differentiated in agent-driven experiences?

Consider agentic commerce as part of a broader digital transformation journey, rather than an isolated AI experiment. The most successful enterprises create cross-functional teams, bringing together data, commerce, customer experience and risk leaders, to align on outcomes and guardrails from day one.

Best practice #4: Start small with hyper-focus on business value

Your strategy sorted, it’s time to think about implementing it. However, instead of aiming for a big splash, consider taking smaller steps through a pilot project or proof-of-concept (POC) that help you get familiar with agentic concepts, apply them to real-life use cases, learn, iterate and scale. 

The first thing to consider is where AI agents can make the most impact: Customer service? Personalization? Dynamic pricing? Process optimization? Define where the business value lies, and go from there. 

That being said, identify focused use cases with clear metrics of success. For instance:

• A customer service agent that handles order status queries.
• A digital assistant that helps shoppers find products by need or occasion.
• An inventory optimization agent that forecasts and updates stock levels in real time.

Once agents consistently deliver measurable impact, scale with intention by adding new capabilities or channels gradually while maintaining governance and control.

Enterprises that succeed treat agentic deployment like product management: Iterative, test-driven and outcome-based.

Best practice #5: Iterate continuously

Launching an agent is just the starting point. Consumer behavior shifts, data patterns evolve and contextual understanding decays over time. That means without ongoing refinement and human oversight, even the best AI agents will drift from their intended goals.

Continuous iteration ensures agents remain accurate, relevant and aligned with both customer needs and business objectives. The most successful enterprises embed improvement loops into their operations.

Here’s what that looks like in practice:

• Monitor real-world interactions to spot degradation, bias or unintentional outcomes. Logs, user feedback and performance dashboards are your early warning systems.
• Retrain and recalibrate frequently using fresh data and updated business rules. Agents should evolve as your assortment, pricing and policies evolve.
• Create human-in-the-loop feedback loops. Just like people, agents need coaching. Assign humans to review agent outputs, correct errors and reinforce desired behaviors. Think of it as digital mentorship with continuous guidance.
• Treat failure as an opportunity. Every customer interaction or agent misstep is an opportunity to improve, not a failure to hide.

When enterprises combine automated iteration with human coaching, agents learn faster, stay compliant and deliver more consistent brand-aligned outcomes. A retail AI assistant that initially misunderstands return policies, for example, can be retrained with updated policy language, turning a potential risk into an improved customer experience.

Best practice #6: Measure impact, not just functionality

It’s tempting to measure success by technical performance: Uptime, API calls or response time. But those metrics don’t prove value.

Instead, focus on outcomes that tie directly to business objectives:

• Adoption: Are agents being used by customers and employees?
• Performance: Are they faster, more accurate and more consistent than previous processes?
• Value: Are they driving incremental sales, retention or savings?
• Satisfaction: Are users reporting better experiences or are frustration signals rising?

A comprehensive measurement framework helps you identify which initiatives to scale, sunset or reinvest in — and builds internal confidence in the agentic roadmap.

Best practice #7: Prioritize security, governance and trust

Agentic commerce introduces new risks around autonomy, data access and brand control. Without strong governance, agents can act in ways that violate compliance or damage reputation.

McKinsey outlines five dimensions of trust for agentic commerce that every enterprise should embed into its governance model:

1. Know your agent (KYA): Verify agent identity (like KYC for humans), require multi-factor authorization for sensitive actions and maintain auditable transaction logs. 
2. Put humans at the center: Personalize based on user-controlled preferences, enable human override for critical decisions, build emotional trust through consistent tone, ethics and empathy. 
3. Embrace transparency: Explain product recommendations, show price comparison, availability and alternatives for validation, etc. 
4. Secure everyone’s data: Use end-to-end encryption for sensitive information, limit data sharing, perform regular security testing and comply with global standards, e.g., GDPR in the EU. 
5. Govern responsibly: Define accountability for agent errors, ensure regulatory compliance (e.g., consumer protection) and establish conflict resolution policies.  

A global bank, for example, built automated logging for every agentic action and required human approval for any financial transaction above a threshold. This combination of transparency and oversight enabled safe innovation without sacrificing trust.

Best practice #8: Augment owned experiences with agentic-based conversational commerce

Agentic commerce is more than shopping on ChatGPT. Business leaders preparing for an agentic future shouldn’t only focus on making their brand discoverable and shoppable on those AI platforms — they must ensure that agentic interactions strengthen, not dilute, their brand. 

While preparing product data for GenAI channels is paramount, enhancing “owned experiences” on the brand’s eCommerce site, mobile app and web chat is equally important. The retailer’s website may have fewer visits or conversions, but it won’t disappear, because the eCommerce site plays a bigger role than discovery → cart creation  → conversion. 

Say a consumer buys a shirt on ChatGPT and, after trying it out, it doesn’t fit. ChatGPT won’t process that product return — the brand will. This is an owned experience, which, as many leading retailers know, should be flawless for the consumer to gauge loyalty and engagement. 

These are the experiences where businesses can (and should) leverage the power of AI agents for their own benefit, such as integrating a customer support agent that not only processes a product but can turn that into a product exchange or even another conversion opportunity.

That means:

• Defining distinctive, owned experiences across the journey (save a sale, protect a relationship, secure the next order) where brand-led agents can deliver value through strategies like conversational commerce.
• Prioritizing owned experiences (apps, websites, loyalty platforms) where you can control the interface, data and outcomes.
• Making your data and offers discoverable through GenAI channels, but ensuring agents always point users back to trusted, branded touchpoints.
• Maintaining brand consistency — tone, service level and values should be recognizable, no matter which platform the agent operates on.

The future of commerce will be a hybrid of owned and embedded experiences. The goal isn’t to pick one over the other — it’s to ensure your brand remains visible and valuable in both.

Best practice #9: Design for brand voice, tone and empathy

In agentic commerce, your AI agents are brand ambassadors. Every interaction, recommendation or response shapes how customers perceive your business. An agent that’s accurate but robotic can erode trust just as quickly as one that makes a factual mistake.

Enterprises leading in this space are intentionally designing their brand voice into their AI systems, ensuring that agents speak, respond and empathize in ways that reflect the company’s values and customer expectations.

Here’s how to get it right:

• Codify your brand tone. Define clear linguistic and emotional guidelines for your agents, from the level of formality to how humor, apology or gratitude should sound. Treat it as part of your design system, not an afterthought.
• Embed empathy and emotional intelligence. Agents should recognize frustration, urgency or delight, and respond appropriately. For instance, when resolving an issue, the language should acknowledge the inconvenience before providing the fix.
• Maintain context across interactions. Empathy is more than tone — and memory is equally important. Ensure your agents can maintain context within a conversation or customer history to create seamless, human-like experiences.
• Continuously test for emotional alignment. Monitor real-world interactions for tone mismatches or tone-deaf responses. Use customer feedback and sentiment analysis as training data to fine-tune behavior.

When brand voice and empathy are built into your agentic strategy, every conversation becomes an opportunity to reinforce trust and differentiate your experience. A returns AI agent that apologizes and offers reassurance in the brand’s friendly tone isn’t just resolving a problem — it’s strengthening loyalty.

Strategy in action: Get started by answering 10 questions

To move beyond experimentation, leadership teams should regularly challenge themselves with questions like:

Strategic & business impact

1. What core business outcomes do we expect agentic commerce to drive?
   (e.g., revenue growth, cost efficiency, personalization, customer retention)
2. Which parts of the customer journey are best suited for AI agents — and which should remain human-led?
3. How will agentic commerce change our competitive position or value proposition?
   (Are we shaping the market or reacting to it?)

Technology & integration

4. Do we have the data, infrastructure and interoperability needed to support autonomous agents across channels?
5. What level of autonomy and decision-making should we allow AI agents to have?
   (Recommendation-only, co-pilot or full transaction execution?)
6. How will these agents integrate with our existing systems, including CRM, eCommerce platforms, supply chain and payments?

Governance, risk & trust

7. What governance and oversight mechanisms ensure transparency, fairness and accountability in AI-driven transactions?
8. How will we protect customer data, privacy and consent in agent-to-agent interactions?
9. What are the potential brand, legal or ethical risks if an AI agent makes a poor or biased decision?

Adoption & change

10. How will we prepare customers and employees to trust and effectively collaborate with AI agents?

The winners in this new era will be those who move early, move responsibly and move with purpose, building agentic ecosystems that not only lift conversions and revenue but also reinforce brand trust and loyalty. 

Don’t navigate the AI momentum alone. Contact our experts to start your agentic commerce now.