AI products
Our first AI product
Invoice Agent
Reads e-mailed, uploaded and photographed supplier invoices, checks every field, and holds each one for a person to approve.
See Invoice AgentAI
Practical AI, built into the systems you run.
We build AI applications the way we build any production software: scoped to a real task, connected to your data, tested, and maintained.
What we build
AI where your work already happens.
Most of the value is in connecting models to the platforms you rely on, not in the model alone. That is the part we specialise in.
- AI agents in Salesforce
- Agents that answer customers, qualify leads and update records inside your org, using Salesforce's own AI features such as Agentforce or a custom model integration.
- AI-ready integrations
- MuleSoft APIs that give language models safe, governed access to ERP, database and CRM data, including Model Context Protocol (MCP) servers.
- Assistants for your documents
- Retrieval-augmented assistants that answer from your policies, manuals and knowledge base, and show the source.
- AI in web and mobile apps
- Natural-language search, recommendations and in-app assistants for React, Flutter and Kotlin products.
- Intelligent automation
- Classifying emails and tickets, extracting data from documents and invoices (as Invoice Agent does), and summarising long records.
- Conversational interfaces
- Chat front ends for customers or staff that hand over to a person when the question needs one.
Our AI products
We run what we build.
Invoice Agent is our own AI product, built the way we describe on this page: one task, checked against real documents, with a person approving every result.
Where it fits
Starting points by team.
Typical first projects. Each one is small enough to prove in weeks and useful enough to keep.
Sales
- Score and qualify inbound leads
- Summarise an account before a call
- Draft follow-up emails from meeting notes
Customer service
- Triage and route new cases
- Suggest replies grounded in the knowledge base
- Summarise long case histories
Operations
- Extract fields from invoices and forms
- Reconcile records across systems
- Flag unusual orders for review
Retail and local commerce
- Search deals in plain language
- Recommend offers by location and interest
- Help merchants write deal descriptions
Retail and local commerce is where we are applying AI to our own apps first: we are exploring natural-language search and offer recommendations for Conch Deals, and writing help for merchants on Conch Business.
How we deliver
From one workflow to a system you can trust.
Pick one task
Start with a single workflow and a measurable outcome, such as time per case or first-response accuracy.
Prepare the data
Identify sources, access rules and privacy constraints before any model is involved.
Prototype and evaluate
Build a working prototype and test it against real examples, so quality is measured, not assumed.
Integrate
Connect it to Salesforce, your apps and APIs, with human review wherever a mistake would be costly.
Run and improve
Monitor answer quality, cost and user feedback, and improve prompts, data and models over time.
Built responsibly
Useful only if it is reliable.
Models and tools
We are model-agnostic and choose per task: commercial APIs such as Claude, GPT and Gemini, or open models hosted in your own cloud when data must stay in-house.
- Grounded
- Answers come from your data and cite it, rather than relying on what a model happens to remember.
- Governed
- Models only see what a user is allowed to see, and we choose providers and settings that do not train on your data.
- Human in the loop
- People approve actions that matter. AI drafts and suggests; your team decides.
- Measured
- Every system ships with an evaluation set, so changes are checked against the same examples.
Questions
What teams usually ask first.
Do we need a lot of data to start?
Usually not. Most business AI uses existing documents and records through retrieval rather than training a new model. What matters more is that the data is accurate and accessible.
Can it work inside our existing Salesforce org?
Yes. We can use Salesforce's built-in AI features or connect an external model through MuleSoft or Apex, depending on your edition, data and requirements.
How do you deal with wrong answers?
By grounding answers in your sources, testing against real examples before launch, showing citations, and keeping a person in the loop for decisions that matter.
Where does our data go?
It stays under your access controls. We pick providers and configurations that do not use your data for training, and can host open models in your own cloud when required.
Build with us
Have a workflow in mind?
Tell us the task, the systems involved and what good looks like. We will tell you honestly whether AI is the right tool for it.