“A job we do by hand, or systems that don't talk to each other.”
Gibson finds the missing piece, builds it around your process and checks it works. Sometimes that is a simple connection, sometimes a purpose-built workflow, and AI only when it is the better fit.
You leave with a clear map of the current work, the expected value, human approvals, risks, test results and the next practical decision. Gibson has taken its own browser telephony workflow into production and has a recruitment workflow in active prototype development.
Last updated 13 October 2026 · By Albert Triolo, Founder and Managing Director
01 · WHAT GIBSON HAS BUILT
Two examples from Gibson's register: one active prototype and one live production workflow. Their status proves practical experience, not that another workflow will achieve the same result, timing or operating profile.
REGISTERED STATUS · WORKING PROTOTYPE
The internal build register describes document parsing, structured candidate scoring, a voice interview path and an operator portal. Its status is BETA and active development. It is presented as prototype evidence, not a production product or customer outcome.
REGISTERED STATUS · PRODUCTION
The internal build register describes a browser phone system as live and operational. This shows that Gibson has taken a clearly defined workflow into production, but it does not prove an unverified cost, speed, reliability or commercial result.
WHAT YOU CAN REVIEW
4These outputs make the proposed problem-solving method and project stage clear before a brief.
They explain the method, not a customer-specific solution or promised result.NEXT PRODUCTION EXAMPLE
OPENA future example will show a real workflow, representative tests, human approvals, live status, known limits and customer permission.
No generic dashboard or stock AI image fills the gap.What you gain: a practical decision and action plan, even when the right answer is to simplify the process or not build AI at all.
02 · HOW WE WORK
Gibson maps the current problem, compares AI with simpler options, runs a focused test when useful and records the evidence needed for the next business decision.
The brief identifies the delay, error, risk, cost or coordination failure, the people affected and the decision that matters. AI is not assumed to be the answer.
Inputs, handoffs, exceptions, systems, volumes and existing attempts are mapped. Sensitive records are not requested through the generic enquiry form.
The workflow names what a person must approve, what the system may suggest or perform, where uncertainty is surfaced and which actions must never happen automatically.
A rule, integration, training change or simpler operating fix may be better than a model. Options are compared against evidence, control, effort and measurable value.
The test states who will use it, which data and examples are included, what a useful result looks like, how it will be checked and when to stop. Everyone can see that it is still a prototype.
The outcome may be to stop, improve the process, run another focused test or plan production as a separate piece of work. A prototype is never treated as approval to go live.
03 · INSPECT THE WORKFLOW
Gibson records the problem, current evidence, simpler options and human decisions first. A focused test is proposed only when it can answer a useful business question.
Named delay, affected users and decision owner.
Inputs, handoffs, systems, volumes and uncertainty.
Simplest viable paths compared before selection.
Low confidence and exceptions return to a person.
Expected behaviour, evidence and stop conditions.
Production remains a separate operating commitment.
Automating an unclear workflow can scale the confusion.
A polished demo can hide weak scenarios and missing data.
Automation without accountable human ownership is not an operating design.
04 · KNOW WHAT STAGE IT IS IN
An assessment produces a decision and action plan. A prototype tests the riskiest assumptions. A live implementation adds production security, monitoring, ownership and ongoing support.
Maps the problem, workflow, risks and options. No working product or automation is implied.
Tests selected examples with agreed data, human checks and stated limits. It is not approved for live business use.
Requires security, reliability, monitoring, access, support, change control and accountable ownership confirmed in writing.
People, inputs, handovers, delays, exceptions, systems and the decision that matters, shown in one view.
Sensitive data, access, human approvals, unacceptable outcomes, uncertainty and who is responsible.
Realistic examples, expected behaviour, useful measures, known limits, reasons to stop and a clear status label.
What the test supports, what remains unknown, the recommended next step, responsibilities and any separate work needed to go live.
05 · IS THIS WORTH TESTING?
The first decision is whether an assessment or focused prototype would be useful. Launch timing, platform and ongoing responsibility are confirmed later and separately.
WHAT SHAPES THE QUOTE
Scope problem, facts, approvals, test and ownershipThe written proposal confirms what Gibson will deliver, the project stage, data limits, testing, approvals, launch target, price and responsibilities.
06 · QUESTIONS BEFORE BUILDING
These answers keep assessment, prototype and live production clearly separated.
An AI consultant should help define the business problem, understand the current work and available facts, compare simpler options, decide what people must still approve and state what a focused assessment or prototype must prove. Recommending AI before that work puts the tool before the problem.
No. Some problems are better addressed through clearer ownership, process changes, rules, training or a conventional integration. Gibson compares those options before recommending AI automation or a prototype.
An AI workflow assessment maps the current process, evidence, constraints, risks, system boundaries and human decisions. It produces an option decision and action plan; it does not claim that a build has occurred or that automation is already approved.
An AI prototype is a focused learning test for selected users, data and situations. Going live requires a separate decision covering security, reliability, privacy, monitoring, access, support, future changes and who is responsible.
A prototype can test whether selected inputs produce useful outputs in realistic situations with agreed human checks. It should expose uncertainty and failure. By itself, it cannot prove that a live system will be reliable, adopted, compliant or commercially successful.
Describe the operating problem, current steps, users, systems, delays or errors, aggregate volumes, existing attempts, required human approvals, sensitive-data constraints and the decision a useful test should support.
No. Use aggregate and non-sensitive context in the generic brief. If a deeper assessment is appropriate, data access, permission, storage, minimisation and deletion responsibilities are agreed before records move.
The workflow names which outputs are suggestions, which actions need review, who can override a result, how low confidence is handled and what is prohibited from automatic execution. The same human-in-the-loop discipline applies to AI and conventional business process automation.
The plan defines representative scenarios, expected behaviour, quality or operating measures, human review, failure categories and stop conditions before testing. Evidence is recorded against those measures rather than relying on a polished demonstration.
Existing systems can be assessed, but compatibility is not assumed. Access, supported interfaces, source and destination fields, credentials, rate limits, failure monitoring and ownership are confirmed before an integration is included.
Ownership, accounts, hosting, data, licences, documentation, support and change responsibility are stated in the written scope. This page does not promise one universal handover or deployment model.
Price and timing depend on workflow breadth, evidence, systems, risk, users, prototype fidelity, evaluation, approvals and production requirements. Gibson confirms the exact assessment, AI prototype or AI implementation boundary in writing before commitment.
07 · CHOOSE THE NEXT STEP
Tell Gibson what is failing, who owns the process, which systems and evidence exist, what must remain under human control and what decision a useful test should support. Albert or the senior team will confirm the appropriate next step.
Gibson Promotions has helped Australian businesses generate revenue for 20 years through traditional media, and now recovers revenue for them through call management, lead reactivation and lead management.