Skip to main content
Arnaldus — Intelligence to Transform

For operations teams

Turn one slow process into a clear AI pilot plan.

For teams handling requests and documents. We work with your process owner and IT team to review the workflow, assess feasibility and agree a practical next step.

Human-reviewed assessment. Scope and fee agreed before work starts.

Intelligence to Transform

A clearer decision before you build.

  1. 01

    Assess

  2. 02

    Pilot

  3. 03

    Operate

Start with an assessment. Pilot and ongoing operation are optional.

Intelligence to Transform

Explore the approach

The first paid engagement

One process. Six useful decisions.

A facilitated assessment with your process owner and IT counterpart. We review the available evidence and record assumptions and gaps in a shared brief.

  • Process brief

    The steps, people, systems and handoffs in one workflow.

  • Baseline KPI

    One metric, the available baseline and any missing measurement data.

  • Feasibility review

    Data needs, constraints and assumptions that need testing.

  • Integration boundary

    The systems and access in scope, with dependencies and exclusions.

  • Human approval points

    Who reviews or approves consequential decisions in a proposed pilot.

  • Pilot recommendation

    A reasoned next step: a scoped pilot, further investigation or no build.

Reviewed with your team

We agree the scope and fee before the assessment begins. A request does not start paid work.

Request a process assessment

A decision at every stage

Assess first. Build only when it fits.

  1. Paid · scope agreed first

    Review one process together.

    Clarify the workflow, baseline, feasibility and approval points in a human-reviewed assessment.

  2. Separately scoped

    Agree a bounded pilot.

    If a pilot is recommended, agree its integrations, acceptance criteria, cost and responsibilities before building.

  3. Optional follow-on

    Plan ongoing operation.

    After reviewing pilot results, decide whether to stop, adjust or agree monitoring, support and change work.

An example to discuss

What could a request-triage assessment explore?

Handling incoming customer requestsIllustrative · not a customer result
  1. Read incoming requests

    Potential AI assistance
    Structure information from messages and forms
    Pattern
    Document & summarise
    Proposed human review
    Customer adviser reviews exceptions
    What to validate
    Required fields and acceptable extraction quality
  2. Classify and route

    Potential AI assistance
    Suggest a category and destination
    Pattern
    Recognise & inspect
    Proposed human review
    Team lead agrees routing rules
    What to validate
    Existing categories, exception handling and system access
  3. Find relevant information

    Potential AI assistance
    Retrieve policies or previous answers with sources
    Pattern
    Advise & find knowledge
    Proposed human review
    Customer adviser checks relevance
    What to validate
    Source quality, access permissions and answer accuracy
  4. Draft a response

    Potential AI assistance
    Prepare a reply for review
    Pattern
    Document & summarise
    Proposed human review
    Customer adviser approves sending
    What to validate
    Review time, tone and a defined quality threshold

An assessment question: Could reading and routing requests be a useful first pilot? Value, effort and feasibility need evidence from your team.

The assessment output

A shared brief for business and IT to review.

Business context

  • Current workflow and bottleneck
  • Baseline KPI and assumptions
  • People and approval responsibilities
  • Pilot decision and acceptance criteria

Human-reviewed process brief

A documented recommendation, reviewed together.

Delivery boundaries

  • Data and integration requirements
  • Access, dependencies and exclusions
  • Risk notes for your own assessment
  • Questions to resolve before a pilot

Public examples

Examples to inform your assessment.

These patterns can help frame a discussion. Whether a pattern fits your workflow depends on your data, constraints and a separate feasibility review.

Illustrative public examples from other organisations, sourced via AI For Good. These are not Arnaldus customer projects or evidence of results from our service.

Recognise & inspect

In your process

Inspections, quality checks and reading documents and photos: invoices, claim photos, forms.

Overstory

Vegetation risk around power lines

What AI does
Analyses satellite imagery to spot trees and vegetation that put power lines at risk.
Who decides
Grid operators and arborists review the findings and decide where to inspect or intervene.

Source: Overstory · Case via AI For Good

Myne

Sorting aluminium scrap by alloy

What AI does
Recognises the alloy of each piece on the belt from sensor and image data, so robots can sort it.
Who decides
People design, run and monitor the process and decide how the recovered metal is used.

Source: Myne · Case via AI For Good

Explore the public examples

Responsible by design

Human review, agreed in the plan.

  • Human control is a delivery principle.

    Before a pilot, we agree who reviews consequential decisions and which approval controls need to be built and tested.

  • Risk notes for your assessment.

    The assessment records known assumptions, data needs and human review points as input for your own risk assessment.

  • Clear about the evidence.

    Public examples and illustrative workflows are labelled. Outcomes for a proposed pilot still need to be tested.

Our approach to trust

Start with a clearer decision.

Tell us about one process that needs attention. We’ll discuss the assessment scope and fee with you before any paid work starts.