AI services for marketing and PR: what our AI actually does — and where it stops
Services
| Service | What AI does | What people do |
|---|---|---|
| AI media monitoring and OSINT | Collects and structures media, opinion, market and competitor signals every day | Analysts verify every conclusion and assess impact; for a Fortune Global 500 energy group this means 730+ daily reports in 24 months |
| AI marketing analytics | Combines channel, campaign, cost and sales data; detects anomalies and measurement errors | Decide what the data means and what to test |
| Budget allocation and forecasts | Prepares budget variants and simulates outcomes as ranges (P50, P80, P90) | Choose the variant and commit the budget |
| Impact testing | Prepares baselines, comparison groups, held-out audiences and regional tests | Judge whether a rise is real impact or coincidence |
| AI branding rules | Checks colours, typography, naming, tone and sources on every material | Approve tone and creative direction |
| Decision tools and cockpits | Refreshes data and shows the age of every number | Own each indicator, its threshold and the action on breach |
| Model council | Five AI models assess the same question independently | Treat disagreement between models as a signal that the case is not obvious, and decide |
| Visibility in AI answers | Tracks how ChatGPT, Gemini, Perplexity and Claude describe a brand | Build the sources and coverage those answers draw on |
Measurement before analysis
The most expensive mistakes do not come from bad analysis but from good analysis on broken data. Before we calculate anything we check whether events are counted once, whether campaign tagging is consistent, whether tracking broke when the website changed, how consent settings affect the data and whether bot traffic is counted as people. Only then do we compare and forecast — always saying what is measured and what is assumed.
Data you can date
Our dashboards do not promise “live data” without proof. Every source shows how old its data is; a source without a timestamp gets the verdict “unknown”, not “fresh”, and a tile without data says “not measured” — never zero. We built this after a client caught a dashboard claiming live data that had stopped updating. We tell this story because an agency that learns from its own mistake is more credible than one that claims to make none.
How we build AI tools
- One file, no external dependencies — a cockpit or calculator opens anywhere, works offline and can be handed over to the client.
- Quality gates before delivery — the base gate has 36 rules in five layers: file integrity, static checks, behaviour, evidence and compliance (AI regulation, data protection, accessibility), plus a human visual check.
- The right model for the task — mechanical work on faster models, decisions and changes on the strongest one.
- Keys and passwords never inside the tool — the client’s key stays with the client.
Dozens of such applications run on Commplace® servers today, built for clients and for internal use.
Where AI stops
AI can analyse, compare, simulate, draft, propose options and flag risks. It cannot, without human approval, send, publish, delete, buy, launch a campaign or make a financial commitment. It does not decide budgets and does not settle which hypothesis is true — that takes a test, not a model’s opinion. Content prepared with AI is labelled where the law and honesty towards the reader require it.
How an AI implementation runs
- Diagnosis — which decision is made in the dark today.
- One screen for one question — a cockpit, not a report, with the source and age of every number.
- Connecting sources — monitoring, publications, web and search data, public statistics and meeting notes in one place.
- Automatic refresh with freshness accounting.
- Model council for high-stakes decisions.
- Handover and care — instructions, backups and a release log. Training is part of the implementation, not a separate product.
We do not sell an off-the-shelf product or a licence, and we do not promise an implementation “in a week”. See also Commplace® AI decision support.
Frequently asked questions
What AI services does Commplace offer?
AI media monitoring and OSINT, AI marketing analytics, budget forecasts as ranges, impact testing, AI branding rules, decision tools and cockpits, a model council and visibility in AI answers.
Is Commplace AI a software product?
No. It is the way the agency works; the tools are built and used as part of client service, not sold as a licence.
Does AI make decisions for clients?
No. AI prepares analysis and options. Decisions about budget, publication and campaigns are made by people.
How do you know AI results are true?
We check the data first, label every number as fact, interpretation, benchmark, hypothesis or to be confirmed, and stop anything without a source and date.
What is a model council?
Five AI models assess the same question independently. Where they disagree, the case is not obvious and a person looks more closely before deciding.
Do you train client teams?
Yes, as part of an implementation: how to read the cockpit, what "not measured" means, when to call a model council and how not to confuse correlation with impact.
More Commplace services in English
- International PR agency — evidence-based public relations for global brands.
- Media monitoring and market intelligence in Poland — media, opinion, market and competitor analysis.
- Crisis communications — readiness, rapid response, recovery.
- Decision cockpits — one screen from signal to action.
- Industrial and B2B marketing — proof for engineers, buyers and boards.
- Market entry strategy — taking a brand abroad, or into Poland.
- Employer branding — a promise the company can keep.
- Website design and branding — brands and websites that lead to a decision.
- Social media and online campaigns — a system, not a stream of posts.
- Commplace® AI — AI decision support for marketing and PR.
- Generative engine optimization — visibility in AI answers.
Talk to us
Commplace® · Liczyrzepy 20, 58-564 Sosnówka, Poland · [email protected] · +48 665 379 071
Author: Sebastian Kopiej, CEO of Commplace®. In public relations since 1996. Data current as of 22 September 2026. Wersja polska