// Commerce AI
AI for ecommerce that respects your catalogue and your margins
NUMVOX builds AI for ecommerce teams on Shopify, WooCommerce, BigCommerce and headless stores. We create recommendations, shopping assistants and merchandising tools that use your real catalogue, stock and margin data, and let your team override the model whenever business judgement says otherwise.
01 / THE WORK
Who this is for.
Off-the-shelf personalisation apps are easy to install and hard to steer. They tend to promote what is easy to compute, such as bestsellers, rather than what fits the shopper, the season and your margins.
Custom ecommerce AI starts from your data and your merchandising rules. The model suggests; your team decides which rules always win, such as excluding low-stock items or promoting a new range.
A fit
- 01Growing online stores with a catalogue large enough that shoppers struggle to find the right product.
- 02Merchandising teams who want AI assistance without losing control of what is promoted.
- 03Brands whose customer service answers the same product questions every day.
Not a fit
- Anyone expecting a promise of higher revenue from an AI feature.
- Projects that are really about ad spend or a theme redesign rather than AI features.
02 / PROBLEM
The problem
Generic recommendation widgets suggest what is easy to compute, not what the catalogue and the margin can support. Shoppers see irrelevant products, out-of-stock items get promoted and the team cannot tell why the model chose them.
03 / APPROACH
Our approach
We connect product, stock, order and behaviour data, then build recommendations and search that combine model scores with explicit merchandising rules. Shopping assistants answer from the catalogue and policies, not the open internet. Every feature is tested against a control group so you can measure its effect honestly.
04 / DELIVERY
What we deliver
- 01Recommendation features tied to catalogue, stock and margin data
- 02Merchandising rules a person can override
- 03Shopping assistants that answer from the catalogue and policies
- 04Review summaries prepared for a moderator
- 05Product description drafting for your team to edit
- 06Private apps or extensions for Shopify or WooCommerce
Not included
- A promise of higher revenue
- Ad spend or a store theme redesign unless it is in scope
- Customer data from sources you are not permitted to use
TECHNOLOGY
Platforms
Shopify API · WooCommerce · BigCommerce
Integrations
Klaviyo · Gorgias · Headless APIs
We use platform APIs and private apps so features work with your existing theme and checkout. Klaviyo integrations bring recommendations into email, and Gorgias integrations let a shopping assistant hand conversations to your support team.
// PROCESS
A visible path from first conversation to delivery
- 01
Review the data
We check catalogue quality, attributes, stock feeds, order history and consent for behavioural data.
- 02
Pick one feature
We choose the feature with the clearest value, such as related products or search, and define how success is measured.
- 03
Build with rules
The model and the merchandising rules are built together, with an admin view for overrides.
- 04
Test against a control
The feature runs for a share of traffic so its effect can be compared with the current experience.
- 05
Roll out and tune
Results guide wider rollout, tuning and the next feature.
Where AI for ecommerce works best
The strongest use cases are product discovery in large catalogues, answering pre-purchase questions about sizing, compatibility or delivery, and helping merchandisers write and organise product content faster.
Data quality decides the outcome. Products with missing attributes, inconsistent naming or poor images limit any model, so improving catalogue data is often part of the first phase and benefits the whole store.
Engagement models
AI feature pilot
One feature, such as recommendations or a shopping assistant, tested against a control group.
Commerce AI programme
Several features built and tuned over time with your ecommerce team.
Augmented engineer
A developer who joins your ecommerce team to build platform extensions and AI features.
Working hours from Lahore (UTC+5)
Our engineers work Pakistan Standard Time (UTC+5). UK, European and Gulf ecommerce teams can review experiments and results with us in a shared afternoon.
For US stores, releases are deployed during quiet overnight hours in American time zones, and test results are summarised in writing before your morning trading review.
Overlap and working arrangements by country: Germany, Saudi Arabia, United Arab Emirates, United Kingdom, United States.
Quality and security
Features are tested on staging stores first and released behind switches so they can be turned off instantly. Recommendations exclude out-of-stock or restricted items by rule.
Behavioural data is only used where shoppers have consented, personal data is minimised and shopping assistants never see payment details.
How pricing works
Cost depends on the feature, catalogue size, data quality and integrations. Running costs include hosting and any model usage, which we estimate from your traffic.
A single-feature pilot with a control group shows whether the investment is worthwhile before you expand.
International clients can see up to 60% savings. Compared with typical US, UK and EU rates for equivalent roles. Actual savings depend on role, seniority and scope.
// PROOF
What you can check before you sign
- Based in
- Lahore, Pakistan (UTC+5, no daylight saving)
- Engagement models
- Project delivery, dedicated teams and staff augmentation
- Ways of working
- Written scope before work starts, and work visible in the agreed repository or environment
- Quality
- Review and testing planned with the scope, not added at the end
Frequently asked questions
Will AI recommendations increase our sales?
We cannot promise that, and nobody honest can. What we can do is build recommendations that use your real catalogue and rules, then test them against your current experience with a control group. That shows whether they improve conversion or order value in your store, so decisions rest on evidence.
Do you build Shopify apps?
Yes. We build private Shopify apps and theme extensions that add AI features to your existing store, using the Shopify API. Private apps are built for your store only. For WooCommerce we build plugins, and for headless stores we integrate directly with your front end and commerce APIs.
Can a shopping assistant answer product questions accurately?
It can answer well when grounded in your catalogue, product attributes and published policies, and when it is told to say it does not know otherwise. We test it on real customer questions before launch and give it a clear handover to your support team, for example through Gorgias.
Can our team control what gets recommended?
Yes, and they should. Merchandising rules such as excluding low-stock items, boosting new ranges or hiding certain categories always override model scores. An admin view shows the rules in force, and changes take effect without a developer, so your team stays in charge of the store.
What data do you need to get started?
Catalogue data with attributes, stock levels, order history and, where shoppers have consented, browsing behaviour. Support transcripts help for shopping assistants. We review data quality first, because missing attributes or inconsistent naming limit any AI feature, and we tell you where improvement would pay off.
// NEXT STEP
Talk through your requirements.
Tell us the workflow, the users, and what you already know. We reply by email after we have read the brief. This form is not a place for passwords or sensitive personal data.
Project enquiry
Discuss E-commerce AI
Share the context a reviewer needs. We follow up by email once the brief is clear.
