The Purchase Conversation: How Communication Science Reshapes eCommerce

There is an opportunity in eCommerce that persuasion literature has pointed to for decades, and that technology has only recently enabled us to act on. The opportunity is the purchase conversation: the guided, adaptive, domain-aware exchange between a customer and a system that understands what they are trying to decide and can assist them in making a good decision.

eCommerce has used chatbots for years. They have improved significantly. Natural language processing is now more advanced. Integration with product catalogues is more thorough. Some systems now manage visual search, multilingual queries, and contextual personalisation. These are true advances. However, they mainly address the information retrieval aspect of a purchase: the customer has a question, and the system provides an answer.

The persuasion literature indicates that information retrieval is one aspect of how purchase decisions develop, and often a smaller one.

How purchase decisions actually form

Petty and Cacioppo's Elaboration Likelihood Model describes two processing routes that operate simultaneously during most purchase decisions. Under central processing, the customer assesses the product on its merits: specifications, ingredients, suitability for their needs, comparison with alternatives. Under peripheral processing, they respond to cues: social proof, brand presentation, the confidence of the recommendation, and how the interaction feels.

A customer evaluating a food product might primarily focus on the ingredient list and nutritional information, while also subtly reacting to how the product is presented, whether the brand appears trustworthy, and if the purchasing process feels straightforward or burdensome. Both approaches influence the final decision. A system that only targets the central route by providing accurate information is still valuable. However, a system that addresses both routes offers significantly greater value, as peripheral cues often determine whether accurate information leads to a purchase.

This is the design space Mitochondria’s Cornea occupies.

What a multi-persona system does

Cornea employs multiple personas within a single conversational experience. For a food and wellness brand, this might include a nutrition specialist who can discuss ingredients, dietary considerations, and health goals with genuine expertise, alongside a product guide who understands the catalogue and can translate a customer's stated preferences into specific product recommendations, and a support persona that handles post-purchase queries, delivery questions, and returns. The customer experiences one conversation, with the system routing between personas based on what the interaction requires at each moment.

The persona architecture relies on Cialdini's research on source credibility. The perceived expertise of the communicator influences how the message is understood. A nutrition question answered by a system that appears as a nutrition expert is processed differently from the same answer given by a generic shopping assistant, even if the information content is identical. The customer's confidence in the answer, and consequently their confidence in the purchase decision derived from it, varies based on the perceived source.

For an electronics retailer, the persona setup might include a product advisor who handles technical specifications and compatibility questions, a comparison specialist who can guide a customer through the differences between similar products with genuine analytical depth, and a service agent who manages warranty queries and repair scheduling. Each persona possesses different domain knowledge, a distinct conversational register, and different guardrails defining what it can and cannot say.

The guardrails are where communication design and governance meet. A nutrition persona operating on a food brand's store stays within the limits of what is factually supportable about health claims. A product advisor on an electronics store does not exaggerate compatibility or make promises about performance that the product cannot deliver. These constraints are set during deployment and define the persona's operational boundaries. The system remains within those limits regardless of how the conversation develops.

The ‘purchase journey’ as a single conversation

The second architectural decision in Cornea is that the entire purchase journey, from first discovery through to completed checkout and subsequent support, takes place within a single conversational thread. The customer progresses through a conversation that adapts to their stage in the decision process.

During discovery, the system asks questions to understand what the customer is looking for. A customer who says "something high in protein that my kids will actually eat" is describing a problem at the intersection of nutritional requirements and taste preferences, and the system reasons across both dimensions to produce a useful response. This is a guided conversation that progressively narrows from broad requirements to specific products.

The relevant products are identified and displayed within the conversation as interactive elements. Product links appear in the chat, allowing the customer to examine them, ask follow-up questions, compare options, and add items to their cart without leaving the conversation. Checkout is initiated from the same thread and redirects to the store's existing payment infrastructure. The conversation manages the decision, while the store handles the rest of the transaction.

After the purchase, the same chat thread remains available for support, including delivery queries, product questions, and returns. The system remembers the context of what was bought and why, so support interactions start with relevant knowledge rather than making the customer re-explain their situation.

This continuity is commercially important because each channel transition in a purchase journey creates friction, and minimising transitions with the payment infrastructure clearly enhances completion rates.

The returning customer

Context retention goes beyond a single session. When a customer returns to a store where Cornea is deployed, the system recognises them and carries forward the relevant history. What they purchased previously, what they asked about, what preferences they expressed, what dietary restrictions or technical requirements they mentioned.

A customer who bought a specific trail mix three weeks ago and returns to the store finds a system that already knows their flavour preferences, dietary restrictions, and the exact configuration they selected. The conversation can start with "would you like the same, or would you like to try something different?" This attention to detail respects the customer's time and shows that the brand remembers them. For replenishment categories (food, personal care, consumables), each subsequent purchase should feel easier than the last.

Ajzen's Theory of Planned Behaviour is applicable here. Perceived behavioural control, which is the customer's belief that they can successfully accomplish their goal, enhances when the system already understands their context. The effort needed to achieve a useful outcome becomes noticeably lower, and less effort leads to higher rates of completion.

Channels and language

Cornea operates on the store's website as a chat widget and on WhatsApp via the Business API. The WhatsApp element changes the nature of the interaction. A customer engaging through WhatsApp is in their own communication environment, on a platform they use for personal conversations, with a notification system that keeps the thread visible. The conversation can start on the website and continue on WhatsApp, or begin directly on WhatsApp. Product links within the chat open the Shopify storefront for cart and checkout. The conversational layer remains on WhatsApp. The transactional layer remains on Shopify. Each platform performs its intended function.

The system is inherently multilingual from deployment, enabling it to process and respond in the customer's language rather than relying on a translation layer applied to English outputs. This distinction is significant because translated responses often contain artefacts that native speakers immediately recognise, which can undermine the perception of domain expertise that the persona architecture aims to establish.

Escalation architecture

The design of what a system will not attempt is just as important as the design of what it will do. Cornea's escalation architecture sets the boundary between what the system handles automatically and what gets routed to a human. Policy questions, product information, purchase guidance, and standard support queries are within the system's scope. Queries outside the configured domain, complaints needing judgement about proper resolution, and situations where the customer explicitly requests a human are all routed to the brand's team with full conversational context transferred.

This boundary is set during deployment and can be adjusted based on what the brand observes in practice. Some brands start with a cautious boundary and expand it as they build confidence. Others begin with wider autonomy and narrow it where the system faces recurring edge cases. The boundary is clear, adaptable, and visible in the system's behaviour. A customer who experiences a transparent, honest handoff to a human team member has a qualitatively different experience from one who encounters a system struggling beyond its limits.

Deployment in days

Mitochondria's typical deployment process involves four phases (Stimuli, Neuroplasticity, Synthesis, Energy) over several weeks, starting with a structured mapping of the client's operational environment. This process is vital for products like Cortex, where the system must grasp deep organisational knowledge, or ATP, where operational workflows are complex and undocumented.

Cornea's standard product, the multi-persona conversational system connected to an existing Shopify catalogue, is deployable within three to six days. The product catalogue is already present in Shopify. The purchasing infrastructure (cart, checkout, payment) is already in place. Cornea offers the conversational intelligence layer that links the customer to this existing infrastructure through a guided, persona-driven experience.

What the deployment requires from the brand is a clear brief on persona configuration (detailing the expertise each persona should represent, the register it should use, and what it should avoid saying), along with access to the Shopify store's catalogue and API. The system connects to the catalogue, the personas are configured, and the widget goes live on the storefront and WhatsApp.

Custom modules, such as product configurators where a customer builds a bespoke product through conversation (like a custom trail mix, a personalised skincare routine, or a configured electronics package), require extra planning and take longer. The standard conversational commerce system works from the existing catalogue and needs minimal configuration time compared to what it delivers.

Where communication science meets product design

There is a background to why Cornea is designed this way that relates to how Mitochondria came into existence.

Before Mitochondria, the founding team ran Brand Content Strategy, a communications consultancy that collaborated with over sixty organisations across twenty sectors on brand development, customer retention, lifecycle marketing, and persuasive communication. BCS spent years within client organisations understanding how purchase decisions are made, how customer relationships grow, and what influences whether an investment in communication technology yields returns.

One pattern consistently emerged. eCommerce brands invested in technological tools (CRM systems, email platforms, chat widgets, analytics tools) that could deliver messages, but they did not invest equivalent thought in the communication architecture that would decide what to say, to whom, in what tone, at what moment, and within what boundaries.

Cornea is the outcome that resulted from this observation. The multi-persona architecture applies what BCS learned about how different communication registers serve different stages of the customer relationship. The guardrail system reflects years of working with brands where the precision of what the system states (and what it refrains from saying) directly influences customer trust. The in-conversation purchase journey reflects an understanding that the fewer steps between intent and transaction, the more of those intents convert.

Petty and Cacioppo's processing routes informed the dual-channel design. Cialdini's source credibility research shaped the persona architecture. Ajzen's perceived behavioural control work determined how the first interaction is designed to feel manageable from the customer's initial message. These frameworks guided the product architecture from the outset, as the people who designed Cornea spent a decade studying how people make decisions before building a system to support those decisions.

What Cornea is for

Cornea offers the greatest value for eCommerce brands where the product requires explanation. This includes specialised food, health and wellness products, customised items, technical equipment, and premium goods; where customers need to understand what they are purchasing and why it suits their needs. Stores where customers can browse, select, and checkout without assistance do not need a conversational commerce layer. However, stores where the difference between browsing and purchasing depends on whether the customer receives guidance during their specific decision-making process are where Cornea delivers measurable results.

The system operates on Shopify stores, across website chat and WhatsApp, in multiple languages. It manages the entire customer journey from discovery to checkout and support. Personas are configured for each brand, with guardrails in place. The standard product can be deployed within days.

The purchase conversation is the most challenging aspect of eCommerce to execute effectively at scale: adaptive, domain-aware, guided, and commercially successful. Communication science provides a clear framework for designing it. The technology now exists to implement it. Cornea brings both together.

—

Mitochondria is an agentic AI product company based in Amsterdam, with operations in Pune. ISO 27001 certified. GDPR- and DPDP Act-compliant by architecture.