Responsible Use AI Policy

At Proactive, we believe technology should empower human creativity, not replace it. For over 30 years, we have successfully navigated and adapted to every major digital shift – from the early web to the mobile revolution. Artificial Intelligence (AI) is simply the next operational evolution.

We embrace AI as a powerful tool for research, administrative efficiency, and structured data analysis. However, we strictly adhere to a Human-in-the-Loop (HITL) methodology. This means that while we use AI to sharpen our spade, a human always digs the hole. We do not outsource our strategic thinking, our corporate soul, or our clients’ competitive advantages to automated algorithms.

We are also cognisant of the impact that AI data centres have, their energy consumption and how this can impact sustainability efforts. To remediate this, as a team, training is provided to reduce the reliance on tools such as LLMs, where more efficient prompt engineering methods can be enabled to generate better, more timely responses. This feeds into our ongoing employee training programme to ensure accountability for our actions and relationships with AI.

Scope: This policy applies to all internal AI activities, including the use of third-party tools such as LLMs (ChatGPT, Gemini, Claude) for content iterations, market research and internal analytics or design tools and inbuilt Photoshop AI features for graphic design and illustration purposes. The suite of tools we use can vary according to role and task. These tools are reviewed periodically for accuracy to ensure they match our expected standards and ensure customer data is protected at all times (see Section 5).

We are committed to full compliance with the EU AI Act and GDPR. As a trusted Irish agency partnering with strictly regulated sectors – including MedTech, Pharma, and Advanced Engineering – we treat data integrity as completely non-negotiable.

• The Transparency Mandate (Article 50): In accordance with the EU AI Act enforcement timeline (effective August 2026), Proactive legally commits to explicit output disclosures. Unless synthetic copy, graphics, or video assets have undergone genuine human editorial review and control (where a human assumes complete legal responsibility), they will be clearly labeled as AI-generated.
• Data Sovereignty & Competitor Leakage Prevention: Standard commercial, consumer-facing versions of public LLMs (such as free or standard tiers of ChatGPT, Gemini, and Claude) default to capturing input text for model retraining. Pasting proprietary client text, data layouts, or engineering schemas into these unvetted windows poses an immediate security threat, as that data can be absorbed into multi-billion-parameter neural networks and inadvertently served as search or prompt responses to competitors. To eliminate this risk, Proactive mandates a blanket ban on public consumer logins. We operate through corporate enterprise tiers, session-isolated team workspaces, or vetted API instances where data training toggles are legally turned OFF.
• Server-Side Stricter Boundaries: Even when training toggles are deactivated, public systems routinely store prompt history for 30+ days on US-based servers. Proactive structures all workflows to ensure no raw Personally Identifiable Information (PII) or unannounced patent layouts ever cross into unsecured server networks.
• Intellectual Property & Copyright Respect: We strictly avoid AI tools trained on non-consensual data harvesting. We perform strict due diligence on all creative vendors to verify their training models insulate our B2B clients from emerging copyright infringement liabilities.

To ensure absolute quality, fact-checked accuracy, and brand safety, Proactive mandates a human check at every stage of work. No asset may transfer from a generative tool directly to a client.

A. Content & Copywriting

• AI Application Boundary: Limited strictly to macro ideation, initial structural outlining, semantic SEO keyword mapping, and summarising heavy technical documentation (e.g., raw engineering specifications).
• The Human Standalone: 100% of final campaign, script, and website copy is written, refined, and contextualised by a human copywriter to align perfectly with the client’s localised tone of voice.
• The Golden Rule: Cutting and pasting raw AI text directly into live client environments without performing due diligence by the user is strictly prohibited.

B. Design & Creative

• AI Application Boundary: Employed safely in sandbox environments for initial mood boards, conceptual video storyboarding, generative canvas extensions (e.g., expanding image boundaries), photo cleanup, and rapid internal prototyping.
• The Human Standalone: Core visual identities, logo marks, typography pairings, and final visual campaign assets are explicitly designed, structured, and signed off by our studio designers.
• The Golden Rule: We do not use AI to generate final brand photography or core structural visual identities; traditional, high-premium human manual craftsmanship remains our unified agency baseline.

C. Digital Marketing & SEO

• AI Application Boundary: Gathering initial competitor research fragments, mapping technical metadata variations, and processing raw analytics tables (Google Analytics 4, SEMrush, AccuRanker) to spot deep performance trends.
• The Human Standalone: All analytical connectors, automated tracking links, and semantic structures are manually built, verified, and cross-examined by the digital team.
• The Golden Rule: Performance trackers and client reporting insights utilising integrated AI models must be run completely in parallel with manual, legacy data collection frameworks for a minimum period of three months to prove absolute statistical integrity before deployment.

D. Project Management & Client Operations

• AI Application Boundary: Processing internal audio transcripts of stakeholder interviews or general team meetings into concise summaries and action items.
• The Human Standalone: Interpreting why data trends occur and consulting face-to-face on what strategic direction to take. High-stakes macro decisions – including media budget allocations, corporate positioning, and legal messaging – remain entirely human-driven.

To guarantee protection for our clients and preserve our 30-year market reputation across Ireland, Proactive maintains a strict policy on the following use cases:

• Deepfakes: Synthetic Personas & Deepfakes: We completely ban the generation of synthetic video avatars or vocal clones of real people or corporate executives without explicit, multi-layered written legal consent.
• Bias-Prone Human Screening: When managing Employer Branding & Talent Acquisition campaigns for scaling brands, Proactive will not design, deploy, or recommend high-risk automated screening architectures for clients without a dedicated high-risk compliance audit. This ensures our pipelines prevent the introduction of demographic or systemic bias, keeping our hiring processes fully equitable, transparent, and human-vetted.

We maintain a live, internally audited Proactive AI Exposure Register. No generative tool is introduced to client workflows without managerial sign-off and a data-privacy verification. This includes:

• Approval Procedures: For tools that are introduced into our workflows, users must champion their cause and justify their inclusion to a line manager to secure approval for usage across the department and/or wider team.
• Subscription Audit: Periodic reviews are conducted to assess the veracity of the output of software tools and their continued use by the team.

We believe transparency builds trust. We treat our AI governance as a “living,” evolving framework that gives our clients total clarity:

• Strategic Scoping: If a project utilises programmatic structures or automated generation models at scale, this workflow will be transparently detailed, scoped, and priced in our initial commercial proposals.
• Total Corporate Liability: Proactive accepts full, unreserved liability for all strategic, creative, and technical work delivered. “The AI hallucinated” or “the algorithm made a mistake” is a defense we do not accept internally, and it is a defense we will never present to a client. The buck always stops with our human directors.

Our AI Policy is reviewed every six months or when material changes to AI use occurs. Changes are logged, approved by the AI Oversight Lead (Aidan Reid or Orla McShane), and communicated internally.

All staff using AI tools must complete annual AI compliance training and acknowledge adherence to this policy. The company’s first step in this respect was mandatory attendance of the Proactive AI Governance Workshop on 22/6/2026 conducted by AI Oversight Team Lead, Aidan Reid.