AI Monitor 2026
What should your company change now?
Page 12 of 124 min readPublished Published by Teklens
Pick one important decision and check whether people and AI agents work by it today. Three practices carry this: Shared Context for one current picture, a Company Brain for clear priorities and a Playbook that writes approved changes back to the source. Only then is it worth accelerating and automating. The study shows that openness is there and adaptability is missing. This page is our interpretation, not a result of the study.
Contents · Short answer
Analysis
The starting point: people are ready, their workflows aren't
Three findings carry this recommendation. Openness to change is the highest score in the study at 3.94. Adapting structures and processes is the lowest at 2.47. And adaptability separates advanced from lagging organisations most sharply (4.13 versus 1.84). Anyone still working on acceptance is working at the wrong end. Anyone who only bolts AI onto existing workflows leaves the foundation unchanged.[Probst 2026, Appendix B]
What people and AI agents need
The advanced profile is clearly ahead of the early stage on data (4.11 versus 2.31), on people and collaboration (4.53 versus 2.73) and on adaptability. We read three needs into this that are the same for people and AI agents: access to current, reliable information, orientation on what matters right now, and the ability to turn what was learnt into new work. An agent without current context starts every task from zero. An agent without a clear priority works on the wrong thing. And an agent whose source is never updated repeats yesterday's mistake tomorrow.[Probst 2026, Table 5]
Three practices: Shared Context, Company Brain, Playbook
- Shared Context (Access). New day, back to zero: people and agents rebuild the context every time because it lives in heads, chats and old files. Response: Connect trusted sources so that everyone starts from the same current picture.
- Company Brain (Orientation). Same picture, different directions: everyone knows the facts but follows different priorities, because goals live in slide decks and no document is binding. Response: Connect daily work to shared goals and decisions and show where it deviates from them.
- Playbook (Adaptability). We learnt something new and still work the old way, because the learning stays in meetings and chats. Response: Approve the change, write it back to the source people and agents work from, and review what works. Learning only becomes adaptability when it changes the work.
Why this is a competitive advantage
Models and tools are open to every company, and they get better every year. What cannot be bought is your own current context, your shared direction and the pace at which learning flows back into the sources. The three work together: Shared Context makes knowledge usable, the Company Brain gives it a direction, and the Playbook keeps both current. Every approved change improves the next decision and every agent that works from that source. The more work agents take on, the stronger this loop becomes, in both directions: an outdated source multiplies mistakes, a current one multiplies good decisions.
From dimension to action
Pick the dimension where your team is most likely stuck. The guide shows the work, the owner, an observable result and a review question. It is an interpretation, not a result of the study.
Study finding
Backing 3.56, resources 3.11, vision 3.09: engagement is there, means and plan are missing.
Practical work
Define one AI initiative with a fixed budget, one accountable person and a result in product language.
Owner
CPO or Head of Product
Observable result
One documented decision with goal, budget and approval that the team can find without asking.
Review question
Was the budget renegotiated in the last three months instead of derived from results?
Study finding
Openness 3.94, talent 3.11, collaboration 3.08: willingness is there, capability lags.
Practical work
Run one workflow from task to approval with the people who will own it later; record goal and priority in the Company Brain so that people and agents work on the same priority.
Owner
Product lead with engineering
Observable result
The second Activity needs fewer handovers than the first, and the team can say why.
Review question
Which question did someone have to answer a second time at the last handover?
Study finding
Architecture 3.22, infrastructure 2.70, integration 2.66: prototypes get built, operation is the bottleneck.
Practical work
For one AI feature, write down the path from prototype to operation: data flow, monitoring, ownership; reproducible environments before the next automation.
Owner
Engineering lead
Observable result
The feature runs for one iteration without manual intervention, and results flow back into the core systems.
Review question
Which integration is maintained by hand today?
Study finding
Access 3.05, governance 2.73, quality 2.61: adoption runs ahead of data maturity.
Practical work
Build shared context: name the sources of a workflow (tickets, decisions, documents, code), assign one person to keep them current and check the sources before an agent works from them.
Owner
Product owner
Observable result
A list of stale specs and open decisions that gets shorter every week.
Review question
Which source did the team's last wrong assumption rest on?
Study finding
Policy 2.67, ecosystem 2.66, EU AI Act 2.63: a systemic gap across all profiles.
Practical work
Record the team's operating constraints for AI: which data to which model, where it runs, who may automate what; make human approval visible in every workflow.
Owner
CTO with legal or compliance
Observable result
A one-page document that every new Activity cites instead of asking the questions again.
Review question
Which decision did the team last delay because a rule was unclear?
Study finding
Adapting structures 2.47, the lowest score in the study and the sharpest divide between the profiles.
Practical work
Change the workflow after every initiative: delete one step, simplify one, reassign one approval; approve the change, write it back to the playbook and the source, and review what works after one iteration.
Owner
Product lead
Observable result
A playbook that has demonstrably changed since the last initiative.
Review question
What changed in roles or workflows after the last rollout?
An application model for software teams
Discover → Define → Build → Operate → Discover is the loop a product team works in. It is not the research model; the six dimensions describe preconditions, the loop describes work.
- Discover
Evaluate customer feedback, competitors and conversations; collect signals.
- Define
Shared refinement with current knowledge; the approved spec attaches to the story.
- Build
Coding with a freely chosen agent; hand over context, bring results back.
- Operate
Connect decisions with results; evaluate support and operations; test the next hypotheses.
Seven steps before anything gets automated
- 1 Clarify the context
- 2 Find the biggest lever
- 3 Delete unnecessary work
- 4 Simplify the workflow
- 5 Record the best playbook
- 6 Accelerate
- 7 Automate
Discuss the biggest lever
30 minutes with a founder. You tell us where your product team suffers most. We tell you honestly whether the AI Product Playbook helps.
How a team gets started
- Pick one important decision from the last three months and the workflow it is meant to change.
- Shared Context: name the sources people and agents work from for it, and who keeps them current.
- Company Brain: record goal, priority and ownership in one sentence and check where current work deviates from it.
- Playbook: delete what does not contribute to the result, simplify the rest, approve the changes, write them back to the source and review what works after one iteration.
- Only now: which steps can be accelerated, which automated, with which approval?
Scope & limits
- The study examines organisations across sectors; the transfer to Shared Context, Company Brain and Playbook is our interpretation.
- The study measures neither the effect of Teklens nor time savings, returns, competitive advantage or legal compliance. We make no such claims here as findings.
- The guide sorts measures by dimension. Which comes first is decided by your context, not by the order of the table.
Questions & answers
Where should we start?
With one important decision, not with a dimension. Check whether people and agents work by it today and start where the team loses the most time searching, reconstructing and updating.
Why does writing back to the source matter so much?
Because people and agents work from the sources, not from meeting notes. If an insight stays in a chat, the next task starts from the old state. Only once the approved change sits in the original system does it change the work the next morning.
Do we need Teklens for this?
No. The three practices and the seven steps work with any tool. Teklens implements them in the Company Brain: connect sources, show deviations from the decision and write approved changes back to the source, so a team can repeat the steps without rebuilding the context every time.
How long does it take?
That depends on the workflow. A first pass with a clearly bounded workflow is possible within a few weeks; we make no 90-day promise.
Sources
The DACH analysis comes from Felix Probst's master's thesis (University of St.Gallen, 2026), which focuses on Germany and Switzerland and was written in collaboration with Studyond and Teklens. The European survey is the Teklens pre-study with 165 participants (research partner for the 2027 edition: ETH Zurich with TUM); Teklens, as publisher, provides the interpretation. Full citations and the locator index: Methodology & sources.
- Probst 2026 · Probst, F. (2026). Capturing AI's Potential: How Ready Are Organisations in the DACH Region? Master's thesis, University of St.Gallen, in cooperation with Teklens. Germany and Switzerland, n = 64, maturity scale 1–5, self-assessment, exploratory.Locators used: Appendix B, Table 5 and Figure 7, Section 5.1
PDF edition
The whole AI Monitor as a PDF
Every chapter, figure and source in one file. Download in exchange for your email address.
We email you the link and occasionally get in touch about the AI Monitor. No sharing, unsubscribe any time. Privacy