Stefan Kløvning

No-code suits a simple, standardized flow built from ready-made connectors.

• Low-code suits a flow built mainly visually but requiring extra logic.
• Full-code suits complex or business-critical integrations where developers build a tailored solution across several systems, with custom rules, security, and failure handling.
• The same tool can function as no-code in one workflow and low-code in another; full-code is not automatically better than no-code.
• The aim is not maximum automation, but avoiding repeated data entry and making systems easier to trust.

A new customer is created in the CRM, then someone copies the name, address, and organization number into the accounting system, re-enters details when the deal is won, and updates the CRM after invoicing.

• Kløvning pinpoints where the process breaks down: someone forgets a field, mistypes, or repeats the same task for the fifth time that day.
• They use that scenario to compare the three options side by side.

Read original

They argue that modern software depends on existing frameworks, dependencies, and code because building everything yourself would be far more expensive and slower, and there is no simple way to remove that exposure entirely.

• What organizations can do is prepare: map and review risks continually, establish backups and alternatives, set up crisis and contingency protocols, and train people to use them.
• Kløvning also warns about lightly engineered, quickly built applications, noting that the same failure hit globally funded platforms.

The post compares the Cloudflare incident with the 2024 CrowdStrike outage, which affected 8.5 million Windows systems. It lists services caught up in the Cloudflare failure, including ChatGPT, Claude, Spotify, X, Uber, and Kløvning's own application Com2.ai.

Read original

First, is the organization ready to invest in the cybersecurity, infrastructure, and oversight that enterprise-ready agentic AI requires?

• Second, is agentic AI the right tool, given that workflow automation, retrieval-augmented generation, prediction models, computer vision, recommendation systems, and OCR each solve specific problems more efficiently and with less risk exposure?
• The promised output is a checklist for matching each technique to the problem it suits, not a verdict on agents.

The two-part series covers what enterprise readiness looks like in practice, then maps the alternatives and provides a checklist for choosing among them.

Read original

Announcing a webinar their company ran for the marketing company Synlighet, they describe its goal as applying AI efficiently while complying with existing and upcoming regulation, and understanding the risks and opportunities of responsible use.

• The content is attributed to legal specialist Legal Counsel Kohári Réka, who has 15 years of legal practice and specializes in ICT law.
• Kohári Réka led the presentation, covering the implications of GDPR and the EU AI Act for corporate AI use, practical aspects of responsible AI, contractual considerations with AI vendors, and cybersecurity measures.

The webinar was delivered for Synlighet. The post offers other organizations a similar webinar or workshop on AI compliance, security, or effective implementation.

Read original

Email, support portals, and phone follow-ups leave the ticket queued and bring generic replies.

• Instead, Kløvning searches for key people, product specialists, or customer success managers at the supplier, then sends a short, direct message explaining the situation and asking for guidance.
• The post treats a professional network as a delivery tool, not just a place to look for a job.

Kløvning says this approach escalated the case, connected them with people who knew the product, and moved the project forward in hours rather than weeks. The post asks readers whether they have used their network the same way.

Read original

Full-code is not automatically better than no-code.

• Avoiding third-party platforms is unrealistic; the answer is mitigation.
• Agentic AI may be the wrong tool even when an organization is ready.
• Kløvning does not promote a single approach; the qualifications show where their judgment lies.

The integration post's rule of thumb is to choose a solution simple enough to maintain but robust enough to handle reality. The outage post argues that single points of failure cannot be cleanly removed.

Read original · Original 2 · Original 3

A newcomer learns the practical judgement behind the choices, not just the vocabulary.

Stefan Kløvning walks through a customer-record scenario: someone creates a new customer in the CRM, then copies the name, address, and organisation number into the accounting system.

• When the deal is won, more details are entered again, and the invoice status has to be updated in the CRM.
• The post identifies the failure points — a forgotten field, a typo, or the fifth time someone does the same job in a day — then sets out three ways to connect the systems.
• No-code fits a simple, standardised workflow using ready-made connectors; low-code fits when those options fall short and parts of the flow need custom logic, such as treating different customer types differently or notifying someone when information is missing; full-code fits complex or business-critical flows where developers build a tailored solution with its own rules, security, and failure handling.
• The same tool can be no-code in one workflow and low-code in another, and full-code is not automatically better than no-code.
• The goal is not maximum automation, but avoiding repeated data entry and making systems easier to trust.

For brands planning a campaign

A collaboration on system integration could offer a decision aid rather than a product pitch: the three-way choice, the trade-off between maintainability and robustness, and a closing question about how many places a customer's details are entered. That structure suits a consultancy offer to review or simplify a client's integration landscape.

That perspective may matter to cautious enterprise buyers weighing whether this is the right view of AI for them.

This post opens an enterprise agentic AI decision with two questions it says matter more than the hype.

• First, is the organisation ready to invest in the cybersecurity, infrastructure, and oversight enterprise-ready agentic AI requires?
• Part one of the series covers what that looks like in practice.
• Second, is agentic AI the right tool at all?
• The post lists workflow automation, retrieval-augmented generation, prediction models, computer vision, recommendation systems, and OCR as options that solve specific problems more efficiently and with less risk exposure.
• Part two maps those alternatives and provides a checklist for matching each to the problem it suits.

For brands planning a campaign

A content collaboration on AI could focus on the checklist rather than a general AI theme: which problems justify an agent, which are better served by simpler automation, and what oversight an enterprise needs before getting started. The two-question frame also gives a brand a natural way to position a readiness assessment or scoping workshop.

Written on the day of a Cloudflare network incident, this post describes downtime affecting major applications such as ChatGPT, Claude, Spotify, X, and Uber, as well as Stefan Kløvning's own application Com2.ai.

• It links the incident to the 2024 CrowdStrike outage, which affected 8.5 million Windows systems, and treats both as reminders of how much software depends on third-party platforms, tools, and code.
• The post notes that these single points of failure are hard to eliminate: building on existing frameworks and dependencies keeps development affordable.
• Without them, development would have cost ten, a hundred, or a thousand times more.
• It ends with four practical preparation steps: map and review risks continually, establish backups and alternative measures, establish crisis and contingency protocols, and train people on those protocols.

For brands planning a campaign

This is a reliability and business-continuity angle rather than a growth angle.

• A vendor of backup, monitoring, or contingency services could collaborate on dependency risk: which third parties a business actually relies on, what the fallback looks like, and how often the plan is rehearsed.
• The point that redundancy is imperfect but mitigable offers an honest framing.

That gives marketers a basis for assessing whether this kind of teaching fits their campaign.

Stefan Kløvning describes a webinar that Nordic AI, where the post says "we at Nordic AI", held for the marketing company Synlighet on compliance and security in AI.

• Its goal was to help ensure AI is applied efficiently and in line with existing and upcoming regulation, while accounting for risks as well as opportunities.
• Legal Counsel Kohári Réka, described as having 15 years of legal practice and a specialisation in ICT law, led the presentation.
• It covered the implications of GDPR and the EU AI Act for corporate AI use, practical aspects of responsible AI, contractual considerations with AI vendors, and cybersecurity measures.
• The post offers a similar webinar or workshop to other organisations.

For brands planning a campaign

This item shows delivered compliance programming and a named legal specialist behind it — useful evidence for a brand assessing AI compliance or governance work. A partnership could adapt the webinar format — GDPR and the EU AI Act in practice, vendor contracts, security measures — for a different audience or sector.

It also gives readers a sense of the day-to-day problems this audience faces, alongside the more structured integration and AI material.

Stefan Kløvning describes using LinkedIn as a work tool in IT operations and project delivery.

• A critical case is stuck with a third-party supplier: the ticket is queued, support replies feel generic, and project progress stalls.
• Instead of the usual email, support portal, and phone follow-up, the post describes searching for key people, product specialists, or customer success managers at the supplier, then sending a short, polite, direct message explaining the situation and asking to be pointed in the right direction.
• The reported result is escalation, contact with people who know the product, and progress often in hours rather than weeks.
• The post asks readers whether they have used their LinkedIn network to unblock a stalled case.

For brands planning a campaign

This is a specific example of practical operations problem-solving, aimed at IT operations and project delivery staff who deal with real workarounds. A collaboration could use the escalation pattern as a discussion prompt or explore vendor escalation paths, support quality, and how teams get unstuck.

Stefan Kløvning says the problem is usually not weak routines but weak connections between systems, then lays out three approaches.

• No-code suits simple, standardized workflows using ready-made connectors, such as creating a customer in the accounting system when they appear in the CRM.
• Low-code suits workflows built mainly visually but requiring extra logic, such as treating customer types differently or alerting someone when information is missing.
• Full-code suits complex or business-critical integrations, with developers building a custom solution across several systems that accounts for their own rules, security, and failure handling.
• The closing argument: full-code is not automatically better.
• The best solution is simple enough to maintain but robust enough for reality; the goal is not maximum automation but avoiding the same data entry five times over.

A repeated pattern: describe a familiar manual workaround, compress the choice into a short rule of thumb, then add a qualifying line to keep the rule from becoming a hierarchy.

Kløvning rejects the idea that more code means more serious work, noting that the same tool can count as no-code in one workflow and low-code in another.

For brands planning a campaign

This is strong material for a vendor selling integration platforms, iPaaS, or ERP/CRM connectors: it gives buyers a way to place their own workflow on a spectrum rather than debate tools. A collaboration could turn the post’s rule of thumb into a practical decision checklist, grounded in workflow characteristics rather than product names.

Examples: Source 1

Kløvning compares it with the 2024 CrowdStrike incident, treating both as reminders of how much software depends on third-party platforms, tools, and code.

• There is no simple way to avoid such incidents entirely: building everything in-house would have been far more expensive and slower.
• The practical advice is to map and continually review risks, establish backups and alternatives, set up crisis protocols and contingency plans, and train people to use them.
• The post also flags the exposure of quickly built, lightly engineered applications when even globally funded platforms fail.

News event, comparison with a prior named incident, then a short numbered set of preparedness actions, ending on the idea that preparation reduces rather than eliminates downside.

For brands planning a campaign

Relevant to vendors in resilience, disaster recovery, backup, and cloud security. The post offers a useful collaboration format: an incident post-mortem that turns a news event into a preparedness checklist rather than a scare story.

Examples: Source 1

First, is a company ready to invest in the cybersecurity, infrastructure, and oversight that enterprise-ready agentic AI requires?

• Second, is agentic AI the right tool for the job, given that workflow automation, retrieval-augmented generation, prediction models, computer vision, recommendation systems, and OCR each solve specific problems more efficiently and with less risk exposure?
• The post says part two maps those alternatives and provides a checklist for matching each one to the problem it fits.

A series format that pairs a readiness question with a tool-choice question and promises a matching checklist rather than a single recommendation.

For brands planning a campaign

Useful for buyers comparing AI approaches or weighing a rushed agentic AI decision. A collaboration could use the fit exercise itself: a short diagnostic that matches a company’s problem to the appropriate technique, backed by the vendor’s implementation experience.

Examples: Source 1

The stated goal was to make sure AI is applied efficiently but in line with existing and upcoming regulation, and that both risks and opportunities for responsible use are understood.

• The presentation was led by legal counsel Kohári Réka, described as having fifteen years of legal practice and a specialization in ICT law, covering the implications of GDPR and the EU AI Act in corporate AI use, practical aspects of responsible AI, contractual considerations with AI vendors, and cybersecurity measures.
• The post closes by offering a similar webinar or workshop on AI compliance, security, or effective implementation to other organizations.

Education-led promotion: describe a session delivered for one company, name what it covered, then offer the same to a wider audience.

This is a company announcement rather than personal commentary. The technical and legal roles are split across colleagues, so ask who would cover each area if the brief requires one named expert to cover both.

For brands planning a campaign

A ready-made format for compliance-focused co-marketing: a live session delivered for a named partner, with an open offer to repeat it.

• A brand selling governance, security, or AI tooling could sponsor or co-host a similar session.
• The legal content is delivered by a separate specialist, not Kløvning alone.

Examples: Source 1

Rather than rely on email, support portals, and phone calls, Kløvning describes searching LinkedIn for key people, product specialists, or customer success managers at the supplier, then sending a short, direct message explaining the situation and asking for the right contact.

• The reported result: escalation, contact with people who know the product, and progress in hours rather than weeks.
• The post ends by asking readers whether they have used their network the same way.

A personal anecdote with a before-and-after outcome, ending with a question to draw comments from peers.

For brands planning a campaign

Less directly relevant to a product pitch, but useful for understanding the operational audience and pragmatic, workaround-friendly tone. Brands targeting IT operations and project managers can use this register—concrete, non-promotional, and focused on unblocking delivery—to earn attention before pitching.

Examples: Source 1

It promotes Nordic AI’s AI compliance and security webinars for Norwegian businesses and invites enquiries for similar sessions. Ask what role Kløvning holds at Nordic AI.

View example

The post does not state their role or whether Com2.ai sits under the same company as Nordic AI. Ask for their role and the relationship between the companies.

View example

A follow-up post advertises comparable webinars and workshops. Ask whether any contract or ongoing arrangement followed.

View example

They describe their own setups in concrete terms: Weis names Claude, ChatGPT, Whisper Flow and BigQuery, and walks through connecting Shopify, Klaviyo and Meta data; Kløvning names the systems in an integration chain and the connectors that join them. Both end posts by asking readers where they stand, keeping the material conversational and comment-driven.

Weis addresses e-commerce growth and marketing execution: email flows, lead ads, Q4 content production, on-site upsells and return windows.

• Kløvning focuses on enterprise software choices, dependency risk, compliance and supplier escalation for IT and business teams rather than growth teams.
• They are also more likely to state a rule and then qualify it.

Read original · Original 2

Nielsen reports that 663 of 798 water utility domains lack a DMARC instruction and that a free mail check ran 12,000 times on 10,850 Danish domains; Kløvning cites the 2024 CrowdStrike outage, which affected 8.5 million Windows systems, when framing Cloudflare's failure. Neither relies on alarm; both end with a question aimed at management.

Nielsen focuses on attackers and access: spoofed mail domains, SPF, DKIM and DMARC, and testing physical controls by walking into a building through reception under Denmark's TIBER-DK programme. Kløvning focuses on software architecture, integration methods, AI compliance and delivery bottlenecks, writing from a supplier and consultancy position rather than that of an adversarial tester.

Read original · Original 2

They treat GDPR and the EU AI Act as design constraints rather than paperwork, and break a broad shift into named parts: Gøtzsche sets out five agent roles with defined outputs, while Kløvning frames a readiness question and lists alternatives such as workflow automation, RAG, OCR and prediction models.

Gøtzsche writes from the commercial leadership seat for CEO, CCO and CGO readers, turning the argument into a taught offer; the posts are architecture sketches for scaling revenue.

• Kløvning's material is more operational: which integration approach to choose, how to prepare for outages and how to get a supplier ticket moving.
• When they address compliance, a legal colleague delivers the teaching.

Read original · Original 2

Both posts focus on implementation.

• One lays out a working setup — Claude, ChatGPT, Whisper Flow and BigQuery, with connectors to Shopify, Figma, Canva, Meta and Drive — and recommends keeping a folder per client with brand DNA, past ads and newsletters, then setting read-only data access before asking anything of the model.
• The other turns that approach into a pre-Q4 checklist: audit every email flow, check whether discount codes still hold, build top-of-funnel before performance spend, and connect Klaviyo and Meta data to review what worked in previous Black Friday and Christmas sales.
• A consultancy selling AI adoption to marketing and e-commerce teams gets teachable practice with named tools and a clear starting order, not architecture diagrams.

These posts focus on practical AI adoption for Danish e-commerce and marketing teams, not enterprise system integration, network or email security, or corporate obligations under the EU AI Act.

• The author describes themselves as not an expert and presents the guidance as pocket tips rather than a structured curriculum.
• For a session requiring named legal authority, ask whether a specialist can provide it.
• This is best suited to the practical adoption half of a collaboration, alongside a consultancy or specialist covering integration and compliance.

Examples: Source 1 · Source 2

Both posts draw on measured findings and end with numbered actions.

• One explains the difference between penetration testing, red teaming and cyber security mystery shopping, illustrating it with an unescorted walk through the front door and a live network socket.
• It cites the National Bank of Denmark's TIBER-DK programme, running since January 2019, as the established version of that testing.
• It closes with three concrete steps, including network access control for unknown devices.
• The other reports that 663 of 798 surveyed Danish water utility domains do not tell recipients to reject mail sent in their name, that a free check has run 12,000 times across 10,850 domains, and that NIS2 has covered drinking water since 1 July 2025.
• It then asks who was responsible for testing and when leadership last saw the result.
• For an audience that values documented findings and audit-style questions, this is the most directly reusable material of the three.

These posts cover security testing and Danish critical infrastructure, not AI compliance or system integration, making this a different collaboration from a session on AI adoption, integration and governance.

• They do not address corporate AI use, GDPR in AI projects or no-code, low-code or full-code integration choices.
• The posts are also Danish and organisation-facing, which narrows the audience a Norwegian software consultancy would reach.
• Use this as a second session or security module, not as a replacement for the integration and AI-governance half.

Examples: Source 1 · Source 2

One post addresses that audience directly, framing compliance as guardrails that help a team move rather than paperwork that slows it.

• It poses two diagnostic questions: whether confidential go-to-market strategy and customer data are sitting in a free tier of a language model because it was convenient, and whether sales teams are ranking CRM leads through automated profiling of people's LinkedIn posts.
• The other outlines a five-agent architecture for a B2B rollout into the German market, with a compliance checkpoint before anything reaches the CRM, and asks whether an AI budget is being spent on the surface or the infrastructure.
• Together, the posts give a consultancy a leadership-level case for why governance work is worth buying.

These posts target decision-makers, not implementers, so they offer less of the integration detail that anchors a practical AI-adoption session for implementation teams.

• They do not include an equivalent of the no-code, low-code and full-code comparison or the CRM-to-accounting duplication scenario.
• The agent example centres on one platform vendor's conference and model release, making it better suited to a keynote than a neutral build guide.
• There is also commercial overlap: one post promotes a paid AI governance course of eight rules and ten short videos, and offers a compliance checklist in return for a message.
• Set the co-hosted session's topic and terms against that existing product rather than assuming it is complementary.

Examples: Source 1 · Source 2