AutoTill: Reinventing eCommerce Checkout as an Agentic Sales Conversation

Walk into any DTC brand's office and ask where the team spends most of its growth budget. The answer is the top of the funnel — Meta, Google, TikTok, creator partnerships, content. Ask where the single most valuable square inches of the business live. The answer is the checkout page. Ask when the checkout page was last meaningfully redesigned. The answer is rarely. Ask what part of the checkout an agent is currently allowed to operate. The answer is none of it.

That asymmetry — millions spent driving traffic to a 1999-shaped form — is the largest unfought battle in eCommerce. AutoTill is built for it.

1. Checkout is the only surface no one rebuilt

Look at the rest of the eCommerce stack and you can see ten years of compounding reinvention. Catalogue: AI search, AI merchandising, AI image. Discovery: TikTok Shop, livestream, creator commerce. Acquisition: Performance Max, Advantage+, Generative ads. Email: Klaviyo flows, lifecycle architecture, AutoEmails. Logistics: same-day, ship-from-store, micro-fulfilment.

Then look at checkout. The dominant pattern is still: collect details, offer one upsell, ask for payment, hope. The brand has no voice in the room at the moment of decision. The agent has no access. The customer is alone with their doubt and a "Place Order" button. This is where the marketing budget goes to die.

2. The checkout is a conversation pretending to be a form

Anyone who has worked a physical till knows what closing actually is. It is a conversation. The customer hesitates; you read the hesitation; you address the specific doubt — sizing, returns, delivery date, gift wrap, "is this the right one?". You recommend the complement they hadn't seen. You confirm the choice. You handle the payment with confidence. Cart-to-cash is a series of micro-judgements made in seconds.

A static checkout form encodes none of this. AutoTill encodes all of it. The agent reads the cart shape, the device, the visit history, the brand context, the time pressure, the abandonment pattern. It opens the right conversation in the right register at the right second. It is the till operator the brand could never afford to put on every transaction.

3. What AutoTill does, in order

• Earns the sale. Confirms the choice in brand voice, surfaces the proof points that matter for this cart (returns, materials, sizing, delivery date), preempts the doubt the data says is most likely.
• Addresses the wobble. Detects hesitation signals — pause, scroll back, field re-edit, browser-tab switch — and intervenes with the targeted reassurance.
• Increases AOV. Suggests the genuine complement the customer would have asked the till operator about. Not the spammy "you might also like" carousel — the one item that finishes the job.
• Recovers the abandon. If the customer leaves, the agent owns the recovery sequence — message, channel, timing, offer — based on the actual abandonment moment, not a generic flow.
• Closes the loop. Logs the entire conversation as a structured signal back to the catalogue, the merchandising and the next cohort's checkout. The agent learns per brand.

4. The pilot numbers

Across the first cohort of UK DTC brands, A/B-tested split deployments showed conversion-rate uplift of +9% to +17% and AOV uplift of +6% to +11% within 60 days, with no upstream creative or paid-media changes. The variation is real — apparel saw the largest conversion uplift, beauty saw the largest AOV uplift, homewares saw the largest abandonment recovery — and the underlying mechanism tracks the brand category exactly as you'd expect.

The number that matters more than either is downstream LTV: the first cohort's repeat-purchase rate at 90 days is 8% to 14% higher than the control cohort's, because the conversation at the first checkout sets a different brand expectation. Customers who were closed by an agent in voice come back into the brand differently from customers who were closed by a form.

5. The category position

Shopify, WooCommerce and BigCommerce are not going anywhere — they are the catalogue and order-management substrate AutoTill sits on top of. The category being created is "agentic checkout", and the buyer is the head of growth, not the head of engineering. The line item on the invoice is incremental revenue closed, billed at an agreed share.

Hyperscalers will not build this. Shopify will not build this in its current form, because outcome-based pricing on closed sales undermines its take-rate model. The challenger window is wide open — and unlike most eCommerce categories, distribution is the real moat: every Creative Marketing ResearchDTC brand we work with is a natural deployment site.

6. What you should do

If you run a UK DTC brand — your checkout conversion rate is the largest single lever you have not pulled. A 10% uplift to checkout is worth more than a 30% uplift to ad creative, because it compounds across the entire spend stack.

If you are a head of growth — the cheapest experiment you can run in 2026 is a 30-day A/B of agentic checkout against your existing one. The downside is bounded; the upside is the entire bottom of the funnel.

To talk through deployment on your brand, book a call. To go deeper on the operating philosophy behind AutoTill and the rest of the portfolio, read Systems, Not Features.