Erik Mashkilleyson

Work / Case study

Avidly AI

Building an agency’s everyday AI capability.

We gathered scattered AI experiments into shared ways of working across the agency.

AI strategy, operating systems and enablement · 2025–

Client
Avidly AI
Project
AI strategy, operating systems and enablement
Year
2025–
Sector
Agency operations, AI
My role
AI strategy · Operating systems · Enablement
Themes
AI, Culture, Transformation
Sheet
04
Fig. 01
Webinar cover: Erno Bärlund and Erik Mashkilleyson among rows of yellow mannequins, with the title AI is boring. Human isn’t.
AI is boring. Human isn’t. The March 2026 webinar with Erno Bärlund.

I lead the meeting point of strategy and AI at Avidly, building the working systems, skills and point of view that turn generative AI into everyday agency capability.

01 Challenge

From experimentation to everyday work

Generative AI creates possibilities across almost every part of an agency. Research, strategy, writing, design and production all offer places to experiment. Making those experiments useful across teams requires shared practices, appropriate context and people who can judge the results.

Since January 2025, I have led Avidly’s strategic AI capability alongside my wider brand and business strategy remit. The work brings together the tools people use, the systems around them and the judgment required to produce work clients can trust.

02 Solution

Giving AI the context to be useful

The foundation includes Avidly’s AI playbook, reusable workflows and shared Claude Projects infrastructure. Finnish and English tone-of-voice systems translate brand principles into working LLM prompts. Together, these give teams a repeatable way to bring the right context into their work.

The practical questions are familiar agency questions. What are we trying to achieve? Who is the audience? What should this brand sound like? What information can we rely on? Who owns the final decision? Making those questions explicit improves how people brief, use and review AI-assisted work.

Fig. 02
Slide, Avoid big technology messes: three levels, from basic tools to custom processes to agents and automation, with untapped potential at the first level and a note not to start at the third.
Brand context, instructions and reusable workflows give each Project a clear purpose.

03 Solution

Avidlyze: the brand as an agent

I am building Avidlyze, an agent that redesigns any Google Slides or PowerPoint deck in Avidly's brand. It reads the content, the visual hierarchy and how the elements relate, then rebuilds the slides.

I designed the architecture and the intent model it works from, built it with OpenAI Codex and Google Antigravity as coding agents, and added validation and regression tests to keep the output reliable.

Fig. 03
The Avidlyze interface in Finnish: three steps, Source, Preview and Brand, with a field for a Google Slides link and an Analyse presentation button.
Avidlyze, test version. Bring in a deck or plain text, check the proposal, get an Avidly version in Google Slides.

04 Solution

Services where AI needs people

The same thinking shapes Avidly's AI services for clients. They follow four steps: inspiration and understanding, insight and exploration, strategy and enablement, and design and creation. Each service states what AI adds and why it is not enough on its own, from AI-assisted brand audits and consumer insight to AI strategy for marketing and fast video production.

Three packages each answer one client question. Tekoälymaine (AI reputation) asks what AI says about the brand, and ends in a situation report and a 90-day plan. AI Playbook asks how the team should use AI, and ends in pilots, a playbook and rollout. Brand AI System asks how the brand stays recognisable, and gives AI the brand's own context, voice, examples and quality criteria.

Fig. 04
Avidly's AI services page in Finnish, headed Tekoäly tarvitsee ihmisen (AI needs a human). Three packages side by side: Tekoälymaine, AI Playbook and Brand AI System, each with its question, contents and outcome.
Three packages, three questions. From Avidly's AI services page.

05 Impact

Building capability through practice

Training is part of the operating system. Every month, Erno Bärlund and I run AI-klinikka for the whole organisation of about 100 people: AI news, AI work from our own projects, and practical tips. The nineteenth session was held in September 2026. It gives everyone a regular place to examine tools, share working methods and discuss the decisions that arise in real projects.

I run tens of AI-assisted projects each year, alongside client AI strategy and foresight work. That practical experience feeds back into the workflows, training and advice we provide.

Fig. 05
Slide, Three questions about AI-generated work: Is your first thought that this was made with AI? Could a competitor make the same thing in five minutes? Does the message miss the point?
Recurring clinics turn individual experiments into shared agency knowledge.

06

A public point of view

My external work connects AI adoption with a wider question: how does human contribution gain value as production becomes more accessible? Talks and sessions span an AI and marketing webinar in 2023, Finnvera training in 2024, the 2026 webinar series, an Oulu session and VERGE: Future of Work.

The March 2026 public webinar with Erno attracted 236 registrations. It provides a public example of the thinking behind the work: human interpretation, taste and context are central to using AI well.

Fig. 06
Slide: The most valuable part comes from people choosing the direction and putting ideas in the right context.
Human ownership and review connect AI capability with professional accountability.

Selected artefacts

Fig. 07
Node-based AI workflow on a canvas: product photos and brand references feed generated photoshoots, static ads, format adaptations and animated versions for a shoe brand.
AI for inspiration.
Fig. 08
Slide in Finnish, AI and brand strategy: AI-generated Heinz ketchup ads show a strong brand amplified, a generic toothpaste visual shows a weak brand's gaps amplified.
AI and brand strategy: AI amplifies what is already in the brand.
Fig. 09
Pencil sketch of a farm landscape above three AI-generated renders: pine trees, a combine harvester and an isometric diorama.
AI in the design process: from sketch to generated elements.
Fig. 10
Hand-drawn map from a house to a destination on grid paper, titled Strategy & Enablement.
Avidly AI: Strategy & Enablement.
Fig. 11
Diagram of a creative process branching and converging across six stages, titled The Evolution.
The Evolution: the creative process with AI and humans.
Fig. 12
Circle of scattered dots with a red path cutting through, titled The Noise.
The Noise: with AI, authentic humanity cuts through.
01What we created
  • AI playbook
  • Reusable workflows
  • Bilingual tone-of-voice systems
  • Shared Claude Projects infrastructure
  • Avidlyze presentation agent (in development)
  • AI clinics and training
  • Client AI strategy and foresight
02My role

Since January 2025 I have led Avidly’s strategic AI capability alongside brand and business strategy.

03Impact
  • Monthly AI clinic for the whole organisation of about 100 people. The 19th was held in September 2026.
  • 236 registrations for the March 2026 webinar.
  • Tens of AI-assisted projects each year.
  • AI playbook, reusable workflows, bilingual tone-of-voice systems and shared Projects infrastructure.
See also