SI & Intelligent Automation

Practical automation. Human judgement.

Super Intelligence (SI) is Datalk's approach to combining intelligent tools, data expertise and human judgement. We help you assess defined tasks, such as finding information, handling documents or supporting routine workflows, and agree how a pilot's results will be checked.

Service 05 / 07

Opportunities to assess

Information becoming useful

  1. Source information

    A defined collection of material

  2. Intelligent assistance

    Extraction, classification and summaries

  3. Human review

    Evidence and interpretation checked by people

  4. Interpreted output

    Measured against defined standards

Source information is handled with intelligent assistance, reviewed by people and turned into an interpreted output that is evaluated against the measures set at the start.

  1. Knowledge access

    Help teams find what they need within a defined collection of material.

  2. Document handling

    Explore extraction, classification and summaries for well-defined information tasks.

  3. Workflow assistance

    Find the routine steps where intelligent support could save effort and improve consistency.

  4. Analytical assistance

    Help organise and explore information, with people reviewing the evidence and the interpretation.

Our approach

Start with a useful outcome.

For organisations exploring intelligent solutions, we help identify a specific task and a practical way to test whether the approach is useful.

The starting point is always specific: a task that takes up hours, information that is hard to reach or a process that needs support. The technology is chosen to fit that need, not the other way round.

A practical route to investment

Pilot first, then decide.

We define the task, look at the information available and agree how success will be measured. A focused pilot then gives you a sound basis for deciding whether, and how, to go further.

That keeps the investment tied to a clear purpose. Human review, access and operational responsibilities are designed in from the start.

Technology

The right model for each task.

Model and deployment choices depend on the task, the information involved and the evaluation criteria. We consider hosted and self-hosted options alongside access requirements and running costs.

  1. Models

    Hosted services for language and document tasks, and suitable open models from Hugging Face that can run through Ollama where self-hosting is required.

    • OpenAI
    • Claude
    • Google Gemini
    • Mistral AI
    • Hugging Face
    • Ollama

  2. Retrieval and workflows

    LangChain and LangGraph for retrieval and workflow steps, with Qdrant for searching relevant material. Outputs still need evaluation and human review.

    • LangChain
    • LangGraph
    • Qdrant

  3. Data science

    Python with pandas, NumPy and scikit-learn for analysis and classic models, PyTorch or TensorFlow when a task needs a model trained on your data, and Jupyter to share the work as it develops.

    • Python
    • pandas
    • NumPy
    • scikit-learn
    • PyTorch
    • TensorFlow
    • Jupyter
See Every Technology We Work With

Other services

Find where SI can help your organisation.

Tell us about the process or information task you would like to improve.

Discuss an SI Use Case