A practical AI playbook for political and communications professionals — three sector-specific frameworks for staying ahead
The real risk from AI for political and communications professionals is not replacement but competitive disadvantage — being "lapped" by colleagues who adopt AI tools faster. Execution speed becomes a visible marker separating AI-adopters from traditionalists. Strategic judgment, voice interpretation, and relationship-building remain irreplaceable; what AI removes is the analysis and production friction that previously consumed the hours needed for that high-value work.
The "Post-it test": before starting any task, ask "Can AI do that?" If yes, delegate the execution and redirect effort to the irreplaceably human decision — what to do with the output, how to interpret it for this client, what the strategic implication is. Human judgment about strategy, relationships, and voice is the constraint; AI removes the friction around it.
The shift: Information synthesis is becoming commoditized. Real-time legislative monitoring tools (e.g., Delve AI) now transcribe speeches and flag relevant sections automatically, compressing hours of analysis into background processing. Stop using AI as a research shortcut; deploy it as the analyst preparing briefings so the strategist focuses on client-specific interpretation and relationship leverage.
Practical action: Build custom prompts around client priority issues before major speeches or events. Target specific policy topics (not broad categories) to flag dollar figures, regulatory signals, and actionable implications. Example from the 2026 State of the Union: memo distribution before official transcripts circulated — time-compression as competitive signal.
The shift: Writing production is becoming automatable, but voice and authenticity remain critical. Generic AI-drafted content erodes client credibility. Building AI infrastructure — custom models, style guides, guardrails — is now core job responsibility, not optional. Unoptimized AI output sounds obviously machine-generated; optimized output is indistinguishable from the client's own voice.
Practical action: Upload 10–15 client materials (prioritizing transcripts alongside written work) into Claude or ChatGPT. Generate a detailed style guide capturing tone, sentence structure, recurring themes, and language the client would avoid. Attach this guide to all future content prompts. Strategic instincts become the premium skill; AI handles the drafting.
The shift: Most voter interaction data — calls, door knocks, events — disappears after collection. AI transcription and analysis of this data reveals real-time constituent trends, enabling individualized follow-up at scale rather than mass broadcasts. Speed without judgment is liability: AI-generated voter contact scripts require stress-testing against district context, candidate voice, and opponent vulnerabilities.
Practical action: Audit existing voter interaction data currently unused (call logs, canvasser notes, event sign-ins, town hall recordings). Identify missing capture systems. Assign ownership of the AI processing pipeline to a specific team member — without accountability, automation won't activate. Trust and transparency become electoral advantages with AI-skeptical voters.
Harbath's playbook documents the practitioner adoption curve that creates the window Kyanos operates in. The lag between AI capability and professional adoption is where Kyanos's value proposition lives: campaigns that haven't yet built their AI infrastructure are the clients who need the AUDIT most urgently, because they don't know what AI is saying about them.
The style guide approach validates the chatbot voice design. Harbath's Playbook 2 — building custom models from 10–15 client materials to capture authentic voice — is exactly the methodology behind the REMEDY brief framework. The chatbot's Bill Campbell persona and the Block 1 voice rules are a formalized version of this: a style constraint derived from the candidate's existing communication patterns, attached to every prompt.
The voter interaction data gap is an AUDIT signal. Harbath documents that most campaign voter interaction data disappears after collection. This data — call logs, canvasser notes — represents the unstructured, unindexed corpus of a campaign's actual constituent intelligence. It is not crawlable. AI systems answering voter questions about "what the campaign stands for" will not have access to it. This reinforces why the AUDIT focuses on what AI can retrieve, not what the campaign knows internally.
"Lapped" is the right framing for early-mover advantage. Harbath's framing — not replacement, but being lapped by faster adopters — is more accurate than "winning" or "losing" the AI race. It suggests a continuous competitive disadvantage that compounds over time rather than a discrete outcome. This is directly applicable to how Kyanos explains urgency to campaigns: a competitor who builds AI legibility now is not just temporarily ahead — they are establishing a structural advantage that is harder to close each cycle.