twimco.Let’s talk ↗︎
Menu
Field notes / Reviews

KODIF review: an agent that can file the claim and ship the replacement. Write the limits first

Most support bots stop at a tracking link. KODIF, now joined with Cavalry, wants its agent to file the claim, start the exchange and ship the replacement inside one conversation. That's useful right up until nobody has written down when it shouldn't.

Editorial engraving of a sunny returns room where a support lead checks a policy sheet at her laptop while a teammate wheels a replacement parcel toward the door.

KODIF is worth a demo for Shopify brands whose after-checkout work is split across a helpdesk, a tracking page, a returns portal and a shipping-protection app, and who are ready to write refund and replacement limits in plain language. As of November 8, 2026 it prices by conversation, from $1 on Starter to $0.75 on Enterprise, billed as a flat annual package. After the November 7 Cavalry deal it aims to run returns, tracking, protection and claims itself. Skip it if you'd rather keep the separate tools you already trust.

What KODIF is, as of this weekend

On Saturday, November 7, KODIF and Cavalry announced they are joining forces to build what they call an agentic post-purchase platform. KODIF built AI agents that read customer intent and reason through store policies. Cavalry built the plumbing after checkout: returns and exchanges, tracking, shipping protection and claims. The announcement came with more funding from Bling Capital, SNR, Mistral and Red Bike Capital, bringing the two companies to $19 million raised in total, and the money is earmarked for joining the two products together and widening what the agent can actually do.

KODIF's own site already describes the combined promise. Support, tracking, delivery protection, claims, returns and warranty sit in one place, and the agent is meant to resolve a request end to end. The example on the homepage is the one every support lead knows: "Where is my package?" The scan has stalled, the agent checks the carrier history, files a protection claim and ships a replacement, all in the same reply.

That is a real change in what a support agent is. Most of the tools brands have tried in the last two years answer questions. This one is built to take the action.

Who it is for

Shopify brands with steady after-checkout volume and too many vendors. If your team runs a helpdesk, a separate tracking page, a returns portal, a shipping-protection app and a warranty form, and a human copies order numbers between them all day, KODIF is aimed squarely at you. The company says its technology is already used by more than 100 consumer brands, and it names household direct-to-consumer labels among them.

It fits less well if you have a returns app and helpdesk you genuinely like, or if most of your volume is presale questions rather than problems after the order. It is also a harder sell for teams that have never written their policies down. KODIF's setup assumes you can say, in a sentence, what should happen when a package shows delivered and the customer says it never came.

Decision guide

One platform, or keep the stack you have

ConsiderationMove to one agentic platformKeep separate tools
Returns and claimsThe agent files the claim and starts the exchange in the same turnThe agent answers, then a person opens the returns or claims portal
PoliciesOne written policy the agent executes, with tool calls you can inspectRules spread across four admin screens and a support macro
ExitLeaving means moving support, returns and protection at onceYou can swap one tool without touching the others
Who it suitsBrands drowning in handoffs between vendorsBrands with a returns app and helpdesk they already like
Decision framework: assess these conditions against your own post-purchase operation.
An agent that can finish the job needs someone to decide which jobs it may finish.

Where it is strong

Policies written like you'd explain them to a new hire. The SOP builder takes plain-language rules and turns them into steps, and KODIF says every tool call is inspectable. The sample policy on its site is a good one to copy in spirit: pull the scan history, file a claim if the last scan is more than 48 hours old, ask for a signed statement above a dollar amount, and always send the new tracking number in the same reply. That last line is the kind of detail that separates a calm customer from a second ticket.

Replaying old tickets before a change goes live. The AI assistant can take a policy change and replay past conversations against it, showing which ones would turn out differently. This is the feature to ask about in the demo. It is the closest thing to a rehearsal an agent with write access can have.

One audit trail. When the agent, the claim and the return live in the same system, the story of what happened to an order is in one place. Anyone who has rebuilt that story from three vendor dashboards knows what that's worth.

Pricing you can read. KODIF publishes its model: one flat annual rate by conversation volume, 24,000 conversations a year on Starter, 75,000 on Growth and 160,000 on Enterprise, with the per-conversation rate falling from $1 to $0.75. A conversation is one ticket worked start to finish. Overage is billed at the same rate, and there are no seat or setup fees.

Where it falls short, or needs your discipline

The merger is a day old. Joining two products takes time, and the announcement says the new money is there to speed up that integration. Ask which actions work today on your stack and which are roadmap.

One platform is also one exit. Consolidating support, returns and protection means that leaving later moves all three at once. Ask how you export policies, conversation history and claim records before you sign.

Paying per conversation, not per outcome. Every ticket the agent works counts, whether or not it resolves it. Intercom's Fin, a common comparison, charges $0.99 per outcome instead. Neither model is wrong, but they reward different things. Read what each counts before you compare the two numbers.

The numbers on the site are the vendor's. The scorecards and percentages on KODIF's pages are its own. Treat them as a picture of what to measure, not a forecast for your store.

Autonomy without limits is a refund leak. An agent that can ship a replacement will ship replacements. The question is whether you wrote down when it shouldn't.

An agent that can finish the job needs someone to decide which jobs it may finish.

Questions to bring to the demo. Which actions can the agent take on our carrier, returns and protection setup today, without a person? Where is the dollar ceiling set, and can it differ by country, product or customer history? What does the agent tell a customer when it hands off, and does the person who picks it up see the whole conversation? How do we pause one action, such as replacements, without switching the agent off? And when a replacement goes out by mistake, who sees it first: us, or the customer's bank?

Sequence of work

A careful first month with an agent that can act

  1. Week 1

    Write the delivered-not-received, damaged and wrong-item policies in plain language, with a dollar ceiling for each.

  2. Week 2

    Replay the last month of tickets against the policies and read every conversation that would change.

  3. Week 3

    Turn on tracking and order-status answers only. Replacements and refunds still route to a person.

  4. Week 4

    Allow replacements under the ceiling. Review a sample of agent-issued replacements every Friday.

Illustrative rollout, not a client screenshot or measured result.

Plain verdict

KODIF is one of the more serious attempts to turn a support agent into something that resolves the problem rather than describes it. Book the demo if your post-purchase work is split across four or five tools and your team spends its day carrying context between them. Go in with three written policies, each with a dollar ceiling, and ask to see them replayed against your own tickets.

Compared with Intercom Fin, KODIF is the bolder bet: it wants to own the returns and claims plumbing too, not just the conversation. Compared with keeping your current stack, it trades flexibility for fewer handoffs. If your returns app and helpdesk work and the pain is small, stay put.

If you want help writing the policies, mapping the handoffs or deciding whether one platform is the right shape for your store, that is the work we do on ecommerce operations. Talk to TWIMCO.

Related work: Paddle8

twimco.

We build alongside the people doing the work.
Bring us the problem you’re thinking about.

Related work

Paddle8 ↗︎