Skip to content

AI consultancy · AI agents for businesses

AI isn't delivered in a demo.

Engineers who sit inside your team, with your data and your legal constraints, until the system works in production: measured, governed and documented.

Talk to an FDE →See the pipeline

Use cases

AI agents for your business.

AI automation where your team already loses hours. We start with the highest-return case and measure it from day one.

  • 01

    Customer service and WhatsApp

    AI chatbots that answer from your data 24/7 and hand over to a person when needed.

  • 02

    Documentation with RAG

    Contracts, manuals and tickets: cited answers over your own sources.

  • 03

    Human resources

    CV screening, onboarding and answers about internal policies.

  • 04

    Data analysis

    Ask your data in plain language and get reports that write themselves.

  • 05

    Operations

    Orders, incidents and repetitive processes automated, with human approval.

  • 06

    Sales

    Lead qualification, proposals and follow-up inside your CRM.

Data in the EU · GDPR and AI Act by design

Pipeline

From discovery to production in 10 weeks.

Every phase leaves verifiable artefacts.

$ kodo discovery --cliente acme

✓ 14 procesos mapeados · 6 fuentes de datos

✓ 3 casos de uso priorizados por ROI

→ atención al cliente · ahorro estimado 38 h/semana

Illustrative figures from a typical project.

  1. 01

    Discovery

    Processes, data and use cases with estimated ROI.

  2. 02

    RAG

    Hybrid retrieval, inherited permissions and citations.

  3. 03

    Guardrails

    PII, jailbreaks, grounding and your own policies.

  4. 04

    Evals

    Custom suites per use case and release thresholds.

  5. 05

    Ship

    Canary releases, observability, runbooks and handover to your team.

Guardrails

Every answer goes through two checks.

Input and output are always validated. If an answer cannot cite its source, it does not go out.

  1. 01 · The person’s question
  2. 02 · Input guardrails
  3. 03 · Retrieval (RAG)
  4. 04 · Model
  5. 05 · Output guardrails
  6. 06 · Answer with citations

  • ■ RAG

    Hybrid retrieval, reranking and verifiable citations over your real sources.

  • ■ Guardrails

    PII, jailbreaks, grounding and business policies on every answer.

  • ■ Agents

    Tools, human approval at critical points and traces of every step.

  • ■ Evals

    Custom suites per use case. If it is not measured, it does not ship.

  • ■ LLMOps

    Observability, cost per query and versioning of prompts and models.

  • ■ On-prem and sovereignty

    Open models in your cloud or data centre when data cannot leave.

FAQ

Questions we always get asked.

What is a forward deployed engineer? +

An engineer who works inside your team, with access to your systems and to the people who use the product, instead of delivering from outside. They build, deploy and stay until it works.

Which models do you work with? +

We are model-agnostic: Claude, GPT, Gemini, Llama or Mistral, in the cloud or on your servers. We choose by evals, not preference.

What happens to our data? +

They never leave where you decide. We work with inherited permissions, anonymisation and deployments on your infrastructure when needed, in line with GDPR and the AI Act.

How long does an AI project take? +

A first use case in production usually takes between 6 and 10 weeks.

Contact

Shall we talk?

Tell us what you want to build, or what isn’t working. We reply within 24 hours.

Book 30 min with an engineer →

or write to c.perera@kodo-360.com