Building Trust Through Action: How AI Shopping Agents Bring DataPal's Vision to Life
From theory to transaction: Our journey building a customer-side AI agent that puts individuals in control of their shopping experience.
We in the DataPal team have been developing our core product for 18 months and have an Alpha Release for closed testing. Throughout this journey, we've emphasised a core principle: fiduciary agents working contractually for the individual, not for other parties [Post: "Meet DataPal and your Personal Fiduciary Agents"]. But what does this actually look like in practice?
Over the past 6 months, we've been building a specific concrete implementation of this vision - an AI-powered shopping assistant that demonstrates how personal agents can radically transform the customer experience. This isn't just another chatbot. It's a working prototype of the agent-to-agent future we've been describing in our thought leadership.
The Intention Economy Arrives
For over a decade, Doc Searls has championed the "Intention Economy"—the idea that true customer relationships should be managed by the customer, not the other way around. Instead of companies guessing what customers want, customers would broadcast their intentions directly:
"I need X by Y date at Z price point."
For years, this remained theoretical. But as Nitin Badjatia notes, we're now living through a "Cambrian explosion" in customer experience—with customer-side AI agents, voice interfaces, and ambient experiences all emerging simultaneously. Standards like IEEE P7012 (MyTerms) are enabling customers to express preferences and intentions, and make them available to organisations under a mutually signed agreement.
Our shopping assistant isn't just a prototype. It's evidence that the intention economy has arrived.
The Problem: Traditional CRM Gets It Backwards
Traditional CRM operates on company assumptions about what customers want, leading to frustration and data privacy concerns [Read: “My personal agents post”]. When you shop online today, you're forced into the company's workflow: browse their website, remember another password, re-enter shipping details, navigate their checkout process, and hope they remember your preferences next time.
Every retailer operates in their own silo, and you're the one doing the integration work. As we explored in "What Does Alice's Data Future Look Like?" , this creates a bleak outlook: inputting details repeatedly, little control over data once given, and constant irrelevant promotional messages.
The Vision: Customer-Side Agents Change Everything
In "My Personal Agents and how they will radically improve CRM and change Customer Experience" , we described how personal agents operate through a mobile app as the customer-managed half of relationships. Relationships do need two parties after all. The shopping assistant we built brings this from theory to reality.
Your Agent Works for You, Not the Store
The assistant connects to your DataPal "Thing I Want" data templates—intentions you've stored about products you're looking for. Maybe you saved "Blue hoodie, size medium, under £40", three months ago. Or maybe you asked DataPal to auto-generate buying requirement outlines – from purchase histories, the renewal dates it can see; or even upcoming birthdays that you’ll need gifts for. The agent remembers and actively works to fulfil that intention.
This is fiduciary agency in action. The AI has one job: help you get what you want, on your terms.
Intelligent Matching
Rather than browsing endless product pages, the agent:
Parses your stated preferences (colours, sizes, price limits, buying window)
Scores every relevant product it can find against your intentions (0-100% match)
Explains why each match is relevant
Presents only the top 3-5 best options, as shaped by your preferences and context
Transparent Identity
When ready to purchase, the agent creates a cart with your DataPal DID (Decentralized Identifier) embedded, aligning with FedID Connect for greater security [Read post: “Meet DataPal and your Personal Fiduciary Agents”]. The retailer knows it's you for shipping, but you control what data they access [Read post: “MyTerms and the Great Online Privacy Re-boot”]. No more creating accounts or passwords.
Automatic Receipt Capture
After purchase, the agent automatically detects payment completion, captures the receipt, and stores it in your DataPal account in structured format. Perfect for expense claims, warranty tracking, or tax reporting - one example of how simple changes in existing data flows can enable data portability [Read post: “Ten Ways in Which People Benefit from Data Portability”].
The Technical Reality: It Actually Works
This isn't vapourware. The system is live and functional:
Backend: NestJS with MCP (Model Context Protocol) for tool orchestration
Beckn Protocol: for standardising expressions of intent and inbound responses
AI Layer: Claude Sonnet 4 with custom tools for product discovery, cart management, and receipt processing
Matching Algorithm: Deterministic scoring system (70% structured matching + 30% AI semantic understanding)
Identity: DataPal DID integration for identity management
Cost-efficient: Uses direct database polling for payment detection instead of expensive LLM API calls
The Path Forward: From Prototype to Platform
This shopping assistant is a proof point, not the end goal. It demonstrates that:
The technology works today: AI agents can handle complex, multi-step workflows reliably
Fiduciary relationships are feasible: The agent serves the individual's interests effectively
Integration is practical: Standard e-commerce APIs (e.g: Shopify) work with agent-driven flows
Cost economics are favourable: Agent automation reduces costs for both individuals and businesses [7]
The market is ready: As Nitin Badjatia observes, "the companies struggling most right now are the ones still trying to maintain control... while customers are ready to just tell them what they want - if only they'd listen"
The next step is extending this approach to any high-value or long-term customer relationship, including banking, insurance, investments, medical services, property management and travel.
But more fundamentally, we need to shift from "how do we manage customers" to "how do we become the obvious choice when customers broadcast their needs" . That's what this shopping assistant enables: customers broadcasting needs, agents responding with value.
The Bigger Picture: Agent-to-Agent Commerce and Beyond
Right now, the shopping assistant talks to standard e-commerce APIs. But imagine when both sides have agents:
Your agent: "I need a blue hoodie, size medium, under £40, delivered by Friday"
Retailer's agent: "We have three options that match. Based on your past purchases, I recommend the fleece-lined one. Current stock: 12 units. Can deliver Thursday."
Your agent: "Confirm purchase. Use stored payment method. Add to my wardrobe inventory."
This agent-to-agent communication ensures secure and auditable data exchanges enabled by a data sharing protocol. The customer agent can interact with multiple potential companies simultaneously, with automated buying based on stated requirements [Read post: “Personal Agents: The Next Evolution In Customer Relationships”].
No browsing. No cart abandonment. No friction. Just two agents negotiating on behalf of their principals, transparently and efficiently.
This vision extends far beyond shopping. Your personal agent becomes your Personal Assistant - all centred on one principle: your data stays yours.
Insurance Management: Your agent stores car details, driving history, and coverage preferences. At renewal time, it automatically requests quotes from multiple insurers, compares based on your priorities, and alerts you 30 days before expiration with better options [Read post: “The High Cost of Customer Data: Why Your CRM Might be a ‘Black Box’ in Disguise”].
Subscription Management: Your agent tracks all subscriptions—Netflix, Spotify, gym membership, software licenses. It sends renewal reminders, identifies unused subscriptions, negotiates better rates, and consolidates receipts.
Utilities & Services: Energy contracts ending? Your agent shops the market based on actual usage. Phone contract renewal? It compares deals against your data history—while never exposing your information until you choose to share.
Healthcare: Your agent maintains medical history, medication schedules, and appointments. It reminds you when prescriptions need refilling and ensures continuity when you switch providers—your health data moves with you [Read post: “The Personal (Digital) Education and Employment Record (PDEER)”].
The Power Rebalancing
For decades, companies have had marketing automation, CRM and sophisticated tech stacks while customers had spreadsheets and frustration. That power imbalance is ending.
The shopping assistant represents this rebalancing in action. Your agent knows what you want, has your preferences and budget, and negotiates on your behalf. Most importantly, it treats your data as belonging to you—because legally and technically, it increasingly will.
This isn't just better CX. It's a fundamental shift from capture-and-hold to co-creation. Companies that embrace this shift will thrive. Those that resist will find customers simply routing around them.
Lessons Learned: What Actually Matters
After several months of development, here's what matters:
Determinism Beats Pure AI for Core Functions: For product matching, rule-based algorithms outperform pure LLM approaches—faster, cheaper, more reliable, and explainable. Use AI for natural language understanding and conversational presentation. Use code for matching logic.
Polling is Surprisingly Effective: You don't need expensive API calls for everything. Simple polling every 10-30 seconds works perfectly for receipt detection at zero cost.
Identity, and specifically identifiers, is the Cornerstone: DataPal DID integration proved essential. Without portable identity, you're back to creating accounts everywhere. With it, you have a single identifier that you control working across all relationships (with appropriate protection against correlation).
Users Deserve Explanation, Not Magic: Surface the agent's reasoning. "This matches 85% because: price ✓, colour ✓, size ✓." Transparency builds trust. Magic creates anxiety.
Conclusion: The Future is Already Here
The technology is working. The standards are now in place. MyTerms/ IEEE7012, and particularly the Intentcasting agreement will make that very real, very soon
We need to shift from "how do we manage customers" to "how do we become the obvious choice when customers broadcast their needs." That's what this shopping assistant enables: customers broadcasting needs, agents responding with value.
This shopping assistant proves the concept works. It demonstrates that the future of customer experience isn't about better CRM—it's about customers having their own agents.
The question isn't whether personal agents will transform commerce. The question is: are you ready to let your agent do your shopping?