Running a small AI-first marketing agency means every tool we pay for has to earn its keep. There is no IT department to approve purchases, no budget padding to hide a €400/month SaaS nobody uses. At Choco Media, our marketing agency tools stack in 2026 is lean, deliberate, and almost entirely AI-augmented. This post is a transparent breakdown of what is in it, what we pay, and what each tool actually does for us day to day.
We reviewed this stack at the start of 2026 and cut three tools that had grown redundant once better AI-native options appeared. What is left covers five functional areas: content production, project and client management, analytics and reporting, paid media, and the AI layer that runs through everything. If you are building or auditing a similar setup, this should give you a realistic starting point — not a wishlist.
One honest note before the list: pricing changes quickly in this space. We have linked to current pricing pages where we can, but always verify before purchasing. Monthly costs below are what we actually pay as of early 2026, rounded to the nearest euro.
Content production tools
Content is the core of our delivery, so this is where the most investment sits. We use a layered approach: an AI model for drafts, a human editor for voice and accuracy, and dedicated tools for media and distribution.
Claude (Anthropic) — €18/month
Our primary writing and reasoning model. We use Claude for long-form drafts, strategy documents, client briefs, and internal research summaries. The quality of extended reasoning is noticeably better than alternatives we have tested for nuanced marketing copy that needs to avoid generic phrasing.
- Used for: blog drafts, campaign strategy docs, client onboarding materials, ad copy variants
- Not used for: final copy without a human pass — the editing layer matters
- Monthly cost: €18 (Claude Pro / Teams)
We pair this with our internal prompt library, which is documented in our post on the exact AI tools we use in production. The prompt library is what makes the output consistent across team members.
Midjourney — €28/month
Visual content for social, ads, and blog featured images where stock photography would look generic. We moved to Midjourney v6 earlier this year and the improvement in photorealistic outputs was significant enough that we use it for a majority of client social imagery.
- Used for: featured images, social media graphics, mood boards for client brand work
- Not used for: product photography or anything requiring brand-specific assets
- Monthly cost: €28 (Basic plan)
CapCut (Pro) — €14/month
Short-form video editing for Reels, TikTok, and Shorts. The AI features — auto-captions, background removal, template application — cut editing time substantially for content that follows a consistent format. We do not use it for brand-level video production, which goes to dedicated editing software.
- Used for: Reels and TikTok content, caption overlays, quick repurposing from longer video
- Monthly cost: €14 (Pro)
Descript — €22/month
Transcription, podcast editing, and longer-form video work. The overdub and word-level editing features mean corrections to recorded content take minutes rather than re-shooting sessions. We also use it to repurpose interview recordings into short clips for paid and organic.
- Used for: podcast transcription, repurposing recorded interviews, team video updates
- Monthly cost: €22 (Creator plan)
The question is never “can we afford this tool?” — it is “does this tool save more hours than it costs?” If the answer is not clearly yes within 60 days, it gets cut.
Project and client management
We documented our content engine setup in our post on building a content engine in Notion, but the full project management picture includes a few additional tools.
Notion — €16/month per seat
Our operational hub. Client deliverables, content calendars, SOPs, brief templates, and internal wikis all live here. We have built automations that populate content briefs from a database of approved topics, and the AI features in Notion are genuinely useful for summarising meeting notes and drafting first-pass SOPs.
- Used for: content calendars, client workspaces, SOPs, brief creation, internal documentation
- Monthly cost: ~€32 for two active seats
Linear — €8/month per seat
Issue and task tracking for anything with a deadline and a dependency chain. Notion is good for documents; Linear is better for tracking who is doing what by when. We use it for website projects, campaign launches, and anything with multiple steps that need to be sequenced.
- Used for: project tracking, deadline management, client deliverable milestones
- Monthly cost: ~€16 for two active seats
Slack — €7.25/month per seat
Client communication and internal async. We keep client channels limited to active retainer clients and archive them when a project closes. We have a lightweight AI bot that surfaces daily priorities each morning from our Linear and Notion queues — not a complex build, but useful.
- Used for: client async communication, internal coordination, automated digest from project tools
- Monthly cost: ~€14.50 for two active seats
Analytics and reporting
We believe in owning your analytics infrastructure rather than renting visibility from platforms. That means GA4 as the baseline, with a few additions for clients who need deeper insight.
Google Analytics 4 — €0
Still the primary source of truth for web traffic and conversion tracking. GA4 has had a rocky transition from Universal Analytics, but in 2026 it is stable and the predictive metrics are genuinely useful for clients with enough traffic volume. We supplement it with server-side tracking for clients where cookie consent significantly limits data capture.
- Used for: traffic analysis, goal tracking, audience building for paid media
- Cost: free
Looker Studio — €0
Reporting dashboards for clients. We template our monthly reports in Looker Studio connected to GA4, Search Console, and Google Ads. The build time per client is now under two hours since we standardised templates. Our AI automation service includes automated report generation for clients who want it.
- Used for: client-facing monthly reports, executive dashboards, campaign performance summaries
- Cost: free
Ahrefs — €99/month
SEO research, competitor analysis, and backlink monitoring. We use the Starter plan and it covers everything we need: keyword research, site audits, content gap analysis, and rank tracking. For GEO work specifically, we use it to identify the structured data and linking patterns that correlate with AI citation.
- Used for: keyword research, content gap analysis, rank tracking, backlink monitoring
- Monthly cost: €99 (Starter)
Hotjar — €32/month
Heatmaps and session recordings for CRO work. We use this on client sites during active conversion optimisation engagements rather than on a continuous basis. The AI summary features added in the last year are useful for surfacing patterns in large volumes of recordings quickly.
- Used for: heatmap analysis, session recordings, form analytics during CRO sprints
- Monthly cost: €32 (Plus)
Paid media tools
Our paid media stack is deliberately minimal. The platforms themselves — Meta Ads Manager, Google Ads — provide most of what we need. We add a few tools for creative production and tracking.
Meta Ads Manager + Google Ads — €0 (platform tools)
Both platforms are free to use; you pay for media only. We run both from native interfaces rather than third-party management tools. The AI features inside Meta Ads — Advantage+ audiences, AI-generated text variations — are mature enough that we use them selectively rather than fighting them. Our paid media service covers full campaign management across both.
Foreplay — €49/month
Creative research and inspiration for ad creative briefs. Foreplay lets us save and organise competitor ads from the Meta Ad Library, and the AI features help identify patterns in creatives that are running for extended periods — a strong signal of performance. This replaced a manual swipe-file process we were maintaining in Notion.
- Used for: creative research, competitor ad tracking, building swipe files for client briefs
- Monthly cost: €49 (Starter)
Triple Whale — €0 (on client accounts)
Attribution and reporting for DTC e-commerce clients. We do not pay for Triple Whale at the agency level; it is billed to client accounts where the volume justifies it. For clients below €20k/month in ad spend, GA4 with server-side events is sufficient.
AI infrastructure and automation
This is the layer that connects everything else and where the most interesting tooling decisions live in 2026.
Make (formerly Integromat) — €16/month
Workflow automation between tools. We use Make for content distribution (publish in Notion, trigger social drafts), reporting automation (GA4 data, Looker Studio update, Slack notification), and client onboarding sequences. It is less technical than n8n but covers 90% of what we need with lower maintenance overhead.
- Used for: cross-tool automation, report generation, content distribution triggers
- Monthly cost: €16 (Core plan)
Perplexity Pro — €18/month
Research and fact-checking with live web access. We use this for competitive research, industry statistics, and verifying claims before they go into client content. The cited sources make it substantially faster than manual research for factual grounding.
- Used for: competitive research, statistics verification, quick market sizing, source gathering
- Monthly cost: €18 (Pro)
OpenAI API — variable (~€20-40/month)
We use the OpenAI API directly for specific automations that need programmatic access: batch content processing, structured data extraction, and a few internal tools we have built. This is not a heavy usage item for us — most AI work goes through Claude’s interface rather than the API.
- Used for: automation scripts, batch processing, internal tools
- Monthly cost: ~€20-40 depending on volume
What we cut in 2026
Transparency means listing what did not make it, not just what did.
Buffer (removed)
We used Buffer for social scheduling until early 2026, when the gap between its AI features and native platform scheduling tools narrowed to the point where it was not worth the cost. Meta and LinkedIn native scheduling handles what we need.
Surfer SEO (removed)
We tried Surfer for content optimisation scoring. In practice, the outputs nudged writers toward keyword density over clarity, which conflicted with our writing approach. We switched to using Ahrefs for research and our own rubric for content structure.
Jasper (removed)
An early AI writing tool in our stack that we replaced entirely with Claude. The quality difference for nuanced marketing copy is significant, and maintaining two AI writing tools was unnecessary.
Total monthly cost
Adding it up: Claude €18, Midjourney €28, CapCut €14, Descript €22, Notion €32, Linear €16, Slack €14.50, Ahrefs €99, Hotjar €32, Foreplay €49, Make €16, Perplexity €18, OpenAI API ~€30. That puts us at approximately €388/month for the full stack, excluding media spend and client-billed tools.
- Content production: ~€82/month
- Project and client management: ~€62.50/month
- Analytics and reporting: ~€131/month (Ahrefs + Hotjar)
- Paid media: ~€49/month
- AI infrastructure: ~€64/month
For a two-person agency delivering full-service campaigns, this is a lean number. The key discipline is treating every subscription as a quarterly renewal decision rather than a default that runs until cancelled.
How to think about your own stack
If you are building a similar setup, the question to ask for each tool is: what specific output does this produce, and could that output come from something we already have? Many agencies accumulate tools that partially overlap because they were added at different times for slightly different reasons. An annual audit usually reveals two or three redundant subscriptions.
The AI layer is also worth revisiting more frequently than the rest of the stack — model quality and tool capabilities are changing fast enough that what was best six months ago may not be best now. We do a lightweight review every quarter and a full stack audit once a year.
If you want to talk through what a tool stack looks like for your specific situation — size, services, output volume — get in touch with us here. We are happy to share what we have learned from building and iterating on this over the past two years.