Automation consulting that starts with the process, not the tool

A consulting engagement turns a vague automation idea into a practical decision: what to automate first, what to leave alone, what access is needed and who owns the next step.

Start a scoping conversation See the delivery method

Automation consulting is a diagnostic and planning engagement for a business that needs to decide where automation is useful before it commits to a tool or a build. Durward maps the current process, tests the practical constraints, compares viable routes and hands the implementation decision back to the person who owns the outcome.

The consulting boundary, in plain language

What is included

Process discovery, stakeholder questions and a practical map of the current route, including the workarounds and exceptions people already manage.
Prioritisation of one or more candidates against frequency, impact, rework, access, data quality and accountable ownership.
A review of native platform capability, configuration, integration and limited-build options before a new AI or automation layer is recommended.
A plan for a controlled pilot or implementation route, with human approval points and acceptance criteria.

What is not included

A promise that every enquiry produces a free audit, a build specification or a recommendation to buy a specific vendor.
Full implementation, training or ongoing operations unless those are separately scoped after the decision is made.
A claim that an automation, provider or deployment option guarantees security, compliance, savings or speed.
Ownership of the client's commercial, professional or customer decisions after the engagement ends.

Four questions the engagement is designed to answer

These are the questions a buyer needs answered before “we need AI” becomes a productive project.

What should we automate first?

We look for a repeated process with a visible trigger, a measurable failure or delay, a named owner and a realistic first boundary.

What should we not automate yet?

A changing process, a missing source of truth, an unowned exception or a high-consequence judgement is a reason to pause, redesign or keep the task human-led.

What information and access are actually needed?

The work names the authoritative source for material fields and starts with the minimum access needed to test the route.

What remains a human decision?

Approval, exception handling, customer-facing judgement and the decision to release or expand the workflow stay explicit rather than being hidden inside a tool choice.

A decision method that produces an implementable next step

The work is structured enough to be useful, but does not pretend that every business needs the same level of discovery.

Clarify the business question

We establish the repeated task, outcome, people involved and the reason the current route is failing or becoming difficult to manage.

Buyer receives: A defined question and the person who will own the decision.
Observe and map the current route

We document the trigger, source records, hand-offs, workarounds and exceptions that must survive any change.

Buyer receives: A process map that shows both the normal route and where it stops.
Prioritise and test feasibility

Candidate work is assessed against volume, rework, data quality, system access, risk and the value of a bounded test.

Buyer receives: A reasoned shortlist, including candidates that should not proceed.
Compare the implementation routes

We assess native capability, configuration, integration, a limited build and, where relevant, a bounded AI-assisted step.

Buyer receives: A recommendation with assumptions, dependencies and a clear boundary.
Agree the pilot or handover

The client chooses whether to proceed, defer or stop; implementation only starts after the owner, test and acceptance condition are clear.

Buyer receives: A practical next-step plan with the owner, test and acceptance condition made clear.

Three consulting moments that make the method tangible

The same method can start with a commercial queue, an operations case or an AI question. These patterns show the diagnostic rather than claiming a live result.

Is an old commercial queue worth recovering?

A sales or service team holds historic quotes, dormant accounts or due-service records and needs to decide whether there is a responsible, repeatable follow-up route.

Trigger: A commercial owner identifies a backlog with no agreed review or ownership process.
Systems or sources: Client-authorised CRM, quote, account or service-date records.
Question tested: Whether records can be deduplicated, sourced, assigned and presented to an owner before any live contact is considered.
Human checkpoint: The commercial or service owner decides the audience, messages, channel and whether the route should ever leave draft-only mode.
Exception route: Suppressed, incomplete, disputed or unowned records become a review issue rather than an automated follow-up.
Success measure: A decision-ready queue and documented reasons to proceed, amend or stop.

Which evidence-preparation step is causing rework?

An operations team repeatedly collects emails, documents or system records before a person can assess an issue or approve the next action.

Trigger: A team can describe a recurring case-preparation task but not a dependable way to gather its evidence.
Systems or sources: Client-authorised documents, messages and operational systems.
Question tested: Whether source extraction, structured preparation and an exception queue can shorten the preparatory work without transferring the professional decision.
Human checkpoint: The responsible operations lead or professional retains assessment, sign-off and follow-up.
Exception route: Missing sources, conflicting records or sensitive material are visible for review and never silently resolved.
Success measure: A source-linked review pack and a clear pilot boundary.

Where can AI help without deciding?

A team has a classification, extraction or drafting task, but is unsure whether an AI component is appropriate or controllable.

Trigger: A repeated unstructured-information task with material variation or a known human bottleneck.
Systems or sources: Approved source material, an existing workflow and the client's access model.
Question tested: Whether the AI task can be bounded, checked against deterministic rules and handed to a person when confidence or context is insufficient.
Human checkpoint: The named process owner approves the task boundary, acceptance check and release decision.
Exception route: A low-confidence output, missing context or changed source route stops for human handling.
Success measure: A feasibility decision with an explicit non-AI alternative where that is better.

Consulting makes responsibilities visible before implementation

The service is useful only if the client can see what they will receive and what they remain responsible for.

Durward prepares

A process map, candidate rationale and a view of the operational constraints that affect the choice.
A comparison of viable implementation routes, including when the existing platform should be used instead of adding a layer.
A pilot or implementation plan with a named owner, data boundary, exception route and acceptance condition.

The client retains

The commercial decision, system ownership, access approvals and choice of whether to proceed.
Approval of customer contact, business policy, exceptions and any decision that affects a customer, case or account.
Ownership of the released operating process; future support is considered separately rather than assumed.

When consulting is the useful first service

Consulting is for a buyer who has a real process problem but needs a reliable decision before committing to implementation.

A good fit

Several possible automation ideas exist, but the team needs a way to choose one with the right boundary.
A repeated process has enough people, records and exception cases to make discovery valuable.
A decision-maker can join the work and take ownership of the pilot, implementation or decision to stop.

Better to start elsewhere

The required answer is a generic software comparison without a defined business process or owner.
The business already knows the exact workflow, systems, scope and acceptance test and simply needs delivery: see business automation services.
The work depends on data or permissions the buyer cannot authorise or explain at a high level.

Follow the route that matches the decision you need

These pages provide the detail after the consulting question has been made specific.

Supporting information: AI Opportunity & Vendor Review: the currently published fixed-scope review · Security and data: our approach to access and information boundaries · Start with a free fit call

Bring the process question, not a tool shortlist

Tell us where the work gets delayed, repeated or difficult to inspect. The first step is to establish whether a controlled automation route is genuinely useful.

Start a scoping conversation See how we work