Skip to content
AtheronLABS

You're visiting from the United States. Prices are shown in US dollars. Not right?

labs@atheron:~/industries/retail/ai$ train agent --grounded

AI for retail service

Agents that answer from your systems, not guesses.

Most retail support is the same few questions asked many times. An agent can answer them from the order, the policy and your size guides, and pass everything else to a person. Because the volume is high and the data is routine, a frontier model through its API is usually the right start.

// what matters here

What is different in this industry.

01

Answers from your systems, not guesses

Order status, tracking and return eligibility come from the commerce platform and the carrier, not from the model's memory.

02

Your policy, enforced in code

Return windows, final-sale items and exchange rules are checked by code before the agent offers anything, so no customer can talk it into a refund the policy does not allow.

03

A person when it matters

Complaints, damaged items and anything outside the policy go to your team with the conversation and the order attached.

04

Busy seasons without a hiring rush

The agent takes the repeat questions during a sale, and your team keeps the conversations that need judgement.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectWeb appAIFrontier model

Orders and returns agent

Hypothetical. Not a client, not a result.

// The problem

A growing online brand's support inbox is mostly order status, returns and sizing questions, and response times stretch every time a promotion goes out.

// What we would build

An agent in chat and email that looks up the order, starts a return within your policy, answers sizing from your own size guides and hands anything unusual to a person with the history attached. It runs on a frontier model through its API, which handles high volume well and is the quickest to start; only what each answer needs is sent.

// What is in it

  • Order lookup and tracking from the commerce platform and carriers
  • Returns and exchanges started within your written policy
  • Sizing and product answers from your own guides, with sources
  • Hand-off to a person with the conversation and order attached
  • Evaluation set of real tickets, run before every change
  • Reporting on what customers ask and what the agent closed

// Stack

  • Frontier model API
  • Python
  • Next.js
  • PostgreSQL with pgvector
  • Zendesk or Gorgias API
  • DigitalOcean

// estimate

Build
≈ US$93,800 to US$143,000, delivered within 21 weeksCAD 133,500 to 204,100
Hosting
≈ US$3,660 a monthCAD 5,205 a month
Support
≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route
Frontier model
Model running cost
≈ US$3,480 a monthCAD 4,955 a month

Timeline by milestone

Discovery
2.9 to 3.4 weeks
Specification and evaluation plan
0.4 weeks
Design approved
2.4 to 3.4 weeks
Core features
1.4 to 1.9 weeks
Full build
0.9 to 1.4 weeks
Working pilot
3.9 to 6.4 weeks
Testing and fixes
0.9 to 1.9 weeks
Launch
0.9 to 1.4 weeks
Production
0.9 to 1.4 weeks

How it is paid

Deposit 20%
CAD 26,700 to 40,820
Discovery 2%
CAD 2,670 to 4,082
Specification and evaluation plan 10.2%
CAD 13,617.00 to 20,818.20
Design approved 2%
CAD 2,670 to 4,082
Core features 6.2%
CAD 8,277.00 to 12,654.20
Full build 4.1%
CAD 5,473.50 to 8,368.10
Working pilot 30.7%
CAD 40,984.50 to 62,658.70
Testing and fixes 2%
CAD 2,670 to 4,082
Launch 2%
CAD 2,670 to 4,082
Production 10.8%
CAD 14,418.00 to 22,042.80
Holdback, 30 days after launch (10%)
CAD 13,350 to 20,410

// questions

Questions we get about this.

Which help desks can a retail agent work in?

Zendesk, Gorgias, Gladly and others with an API, or your own chat widget. It replies where your team already works.

What shopper data goes to the model provider?

Only what each answer needs: the order and the question. Payment details never do. If that is still too much, an open-source model in your own cloud is priced as an option.

Can it recommend products?

It can suggest items from your catalogue for a size or an occasion, using only products in stock. It never invents a product or a price.

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

AI customer service agents for retail and e-commerce | Atheron Network Labs