AI at Work: What Has Changed? Part 2

In this article, six experts — a product marketing consultant, a founder and product builder, a business development rep, an interim HR lead, the co-founder of a mobile app builder and a head of marketing — answered the same questions about how AI is changing their everyday work.

Elena Avramenko, co-founder of Modaal, built and released an app, the one-second VHS diary ONSE, in two weeks “with no help from other engineers,” as a “non-tech person.” Elena “expected that building faster would be the main advantage.” Ariadna Batlle, head of marketing at Dost, had expected AI to make data analysis and reporting “much more straightforward.” Both found that speed alone wasn’t the point. For Elena, the hard part is no longer building but awareness, distribution and standing out among look-alike products. For Ariadna, faster doesn’t necessarily mean better, especially when the data behind AI’s answers isn’t reliable.

Sharon Groenewold, an interim HR business partner at tech and deeptech scale-ups, uses AI as a second brain and “created a design system in Claude Design that automatically applies our brand guidelines to presentations.” At the companies Sharon works for, she uses general AI tools, especially ones the IT department has approved. In her view, worries about leaking company information and uncertainty about how AI works keep people from using AI to its full potential.

The other three experts talk about trust directly. Simon Schmitz, who runs the product marketing consultancy Lemoncore, built a tool with Claude for clients’ marketing copy and isn’t sure how much founders will trust it with their voice. For Amine Benjelloun, co-founder and CTO of Mailwarm, “things that previously required several people and several days can sometimes be tested in a few hours,” and Amine thinks trust will shape how much access AI gets. Rylan Cole, a BDR at Corgi, now has AI following up on deals and trusts it to auto-reply, but only in specific scenarios, something Rylan didn’t expect back in 2023. Still, Rylan feels AI washes out a little of the personal charm that matters in sales.

Across these answers, we see a shared theme: AI is now part of the work for all six, and each of them is weighing how far to trust it and what to keep for themselves.

Most sections are marked What they do when they describe the experts’ own practice, or What they think when they share a view or a forecast.

What they have in common

We studied the six sets of answers side by side. Six themes came up more than once: some in what the experts do, others in what they think.

What they think

How far to trust AI

Reviewing the agent’s plan, validating the numbers, making AI’s work easier to verify, letting AI auto-reply in specific cases, worries about company data: each of them has something to say about checking AI’s work or trusting it.

6 of 6

What they do

Making things with AI

An app, a client tool, dashboards, a design system and new software products, built with AI.

5 of 6

What they think

The hard part moved

AI speeds up the making. The hard part is now deciding what to build, how to stand out, what to say and what to measure.

4 of 6

What they do

Everything in one place

A single source of truth for the data tools, a Claude project for every part of the work, a second brain, Claude connected to almost every daily tool.

4 of 6

What they think

There is a catch

With fragmented data, answers can look right but may not be reliable. Fuzzy thinking scales. A wrong decision becomes a finished feature. Easier can also mean lazier.

4 of 6

What they do

Roles change

People work more on their own or outside their original role. One team dropped sprint planning and large backlogs.

2 of 6

What they made with AI

What they doAriadna · Simon · Elena · Amine · Sharon

Five of the six experts describe something they made with AI: an app, a client tool, dashboards, a design system and new software products. One expert described a task from outside work, and it made the list too: Rylan found an apartment in a day with Claude’s help.

Ariadna Batlle

Head of Marketing, Dost

Data spread across platforms

One source of truth

Dashboards that connect the funnel, from channel performance and lead generation to pipeline and revenue.

Numbers validatedLess manual reporting

Elena Avramenko

Co-founder, Modaal

Two weeks, no other engineers

An app called ONSE

A one-second VHS diary: record one second of your day, every day, and the app combines those seconds into a VHS-style video.

A spec for each featureThe agent’s plan reviewedThen the code

Simon Schmitz

Founder, Lemoncore

Clients didn’t build the homepage

The Messaging Studio

Clients find their workshop messaging inside a prompt builder and write ads, emails, posts and web copy with it.

Claude wrote the codeGitHub and Netlify

Amine Benjelloun

Co-founder & CTO, Mailwarm

From a rough idea

Something users can try

AI researches the market, challenges the idea and helps write the spec. Coding agents build a large part of the product.

DebuggingLanding pageLaunch content

Sharon Groenewold

Interim HR Business Partner / HR Lead, MeaningfulDot

Brand guidelines

A design system

Built in Claude Design, it applies the brand guidelines to presentations automatically.

Claude Design

Rylan Cole

BDR, Corgi

Outside work

An apartment in a day

Close to work, within a budget, and fast. Claude filtered every rental listing in a radius by Rylan’s preferences.

Lease signed within a day

Simon’s “old deliverable was a wireframed homepage with all the copy written. Clients liked it. Then they didn’t build it. That told me a file was the wrong handoff.” So Simon used Claude to build the Messaging Studio. “I’m a GitHub beginner,” Simon says. “… Claude wrote the code and walked me through the hosting.”

Elena, who discovered vibecoding in April 2025, built ONSE through specs: “For each feature I write a short spec, review the plan the agent proposes, and only then let it write code.” For design, Elena used Claude, then moved the finished design to Modaal.

For Amine, AI now takes part in almost every step of a new product, from the first idea to something users can try: “this is no longer just ‘AI helping me write code.’”

Simon’s first version of the Studio worked, “but the prompts were too generic. Now I’m rebuilding them around real situations instead of content types.” Ariadna worked with AI iteratively on the dashboard logic, the calculations and the gaps in the data, and kept one step: “I still validated the numbers and the underlying assumptions, because having a dashboard is only valuable if the data behind it is reliable.”

The hard part moved

What they thinkElena · Amine · Simon · Ariadna

In Elena’s words:

Three other experts point upstream: Amine to deciding what is worth building, Simon to knowing what to say, and Ariadna to clearly defined metrics and data that is ready before AI works with it.

What hasn’t changed is deciding what is worth building. AI can generate 100 solutions very quickly, but choosing the right problem, understanding users, having taste, distribution, and knowing when something is actually working are still very human problems.

I also think cheap writing moves the hard part upstream. … When anyone can produce copy, the scarce thing is knowing what to say.

For Simon, the core of the work stays the same: “My workshops are still me and a founder on a Miro board, arguing about who their ICP really is. AI is in the room now but mostly for note taking.”

Everything in one place

What they doAriadna · Simon · Sharon · Rylan

Four experts keep their work in one place with AI: a single source of truth at Dost that connects the data tools, a Claude project for every part of Simon’s work, Sharon’s second brain, and Rylan’s Claude, connected to virtually every tool Rylan uses day to day.

We have one single source of truth that connects all different data tools: spreadsheets, CRM + Slack + Notion, so we can process and follow information from one single place.

Nowadays, I have a Claude project for every single part of my work. … It also knows my frameworks and my rules, which can be really helpful when trying to brainstorm ideas.

I use it currently as my second brain; to brainstorm, structure information of meetings, send reminders etc.

I have Claude connected to virtually everything I use on a day-to-day basis.

“I’d say the main thing is pipeline management,” Rylan says. “Before AI, I’d spend hours going through my deals and tracking timelines etc. Now I have Claude automatically notifying me when a deal is going stale or when someone is on track to close. It’s great.” At Dost, Ariadna’s team also has “admin / logistic agents that help with event coordination” and “SEO/GEO agents that help with content ideas and content creation.”

Sharon names the other side of this.

There is a catch

What they thinkSimon · Ariadna · Rylan · Elena

Four experts describe a catch in how fast and easy AI makes things.

I realized that if your data is fragmented, your tracking isn’t consistent, or your metrics aren’t clearly defined, AI can actually make things more confusing by giving you answers that look right but aren’t necessarily reliable.

I think what I underestimated initially is that AI enables you to turn ideas into something tangible you could potentially sell to clients at the speed of light. That makes it really enticing to just try out new ideas, … which can be a double-edged sword. It can lead you down the wrong path really quickly if you’re not careful, but it can also let you execute a lot faster and find your right product-market fit a lot quicker than you used to in the old days.

AI can make your life easier, but also lazier.

For Elena, the catch is the one quoted above: with an agent that implements a spec within hours, a wrong decision “turns into a finished feature that someone then has to remove.”

Amine describes a different kind of speed: not the speed of the work, but the speed at which AI itself evolves.

Who does what now

What they doAmine · Elena

Two experts describe roles changing inside their own teams.

I think collaboration is moving from producing things together to reviewing and directing things together.

AI has changed the role of everyone on the team and made each of us much more autonomous. We now do daily calls mainly to catch up, share what we’ve done, and make decisions.

We’ve also removed things like sprint planning and large backlogs because execution has become so much faster. Planning too far ahead often doesn’t make sense anymore.

Elena sees it at Modaal: “Each person on the team now does work outside their original role. My tech co-founder builds prototypes and records YouTube Shorts. My designer writes Help Center articles and does simple integrations. Roles are mixed and borders are blurred.” At a time when it is unclear who does what, Elena says, having people with high agency makes the team “move 10x faster.”

The other four experts describe what changed with colleagues and clients.

Sharon Groenewold

Interim HR Business Partner / HR Lead, MeaningfulDot

In meetings

Summaries and follow-ups

Transcription tools summarise meetings and share follow-up actions. AI is also coming into the core HR systems Sharon works with, for example chat functions.

Ariadna Batlle

Head of Marketing, Dost

With sales and product

A clearer picture

Marketing, sales and product often see growth differently. AI helps bring the data together into a clearer picture of the funnel.

Simon Schmitz

Founder, Lemoncore

With clients

Where the work ends

“Before, it ended when I sent a document. Now the messaging keeps working after the engagement …”

Rylan Cole

BDR, Corgi

With colleagues and clients

Emails

“All I really need to say is emails. I think most will resonate with that.”

How far they trust it

What they thinkAmine · Ariadna · Rylan · Simon · Elena · Sharon

All six experts say something about checking AI’s work or trusting it, from different angles. Elena reviews the agent’s plan before any code is written. Ariadna validated the numbers behind the dashboards. Rylan’s trust has grown since 2023.

I initially used AI back in 2023 during college to simply fix grammar errors and fluff my essays. Never did I think I’d have it following up for me on deals, let alone trust it to auto reply to specific scenarios.

I’m still unsure how much founders will trust a tool with their voice. My guess is they’ll want a human to check it for a while.

Trust is the piece that I think will change how much access we give to AI.

Today we still verify a lot of the work AI does, and that becomes difficult as we give agents larger and more complex tasks. I believe the tools that allow us to trust AI more by making its work easier to verify will be extremely important.

Sharon looks at it from the people side, as part of AI adoption in a company.

What they expect next

What they thinkAriadna · Amine · Sharon · Elena · Simon · Rylan

NoteThese are expectations, not facts, and they point in different directions.

Three experts see agents taking on more of the work, each with a different degree of certainty: Ariadna says agents “could take a step further,” Amine “can imagine” teams of humans and agents, and Sharon expects AI HR agents to become “colleagues.”

AI agents could take a step further: monitoring performance across channels, identifying opportunities or bottlenecks, and suggesting actions based on what’s actually driving pipeline and revenue.

I can imagine teams having a mix of humans and agents. You assign a task to a colleague or to an agent in almost the same way. The agent has access to the right context and tools, does the work, and comes back when it needs a decision or approval.

The full admin load will be automated, AI HR agents will become our ‘colleagues’ when it is about the practical aspects of HR and first questions from employees.

Amine is still figuring out “how memory, permissions, trust, and collaboration should work in that world.” Elena’s forecast goes beyond one team, to what kinds of companies get built.

I expect more personalised products and many small lifestyle businesses, and fewer new large companies, possibly none. I expect product, design and engineering to merge into “builders” who lean technical, design or research. I’m unsure whether large new companies can still form when building is this cheap.

Simon wants the Studio “to help before the sale, too,” with a score for how well each prospect fits.

Rylan sees it differently:

The tools they named

What they doRylan · Ariadna · Sharon · Simon · Elena · Amine

All six experts named Claude. Three of them also use ChatGPT, and two use Lovable.

Rylan Cole

BDR, Corgi

Automation, emails

Claude

Design, or when Claude credits run out

ChatGPT

Ariadna Batlle

Head of Marketing, Dost

Tools

ClaudeChatGPT

Agents at Dost

Event coordinationSEO/GEO content

Sharon Groenewold

Interim HR Business Partner / HR Lead, MeaningfulDot

At client companies, approved by IT

Mainly Claude, so far

Presentations

Claude Design

Own company

LovableFireflies

Simon Schmitz

Founder, Lemoncore

Most of the work

ClaudeProjectsCustom skills

Calls

Granola

LinkedIn graphics

ClaudeFigma

Workshops

MiroMural

Hosting the Studio

GitHubNetlify

Elena Avramenko

Co-founder, Modaal

Drafting, research, strategy, design

ClaudeClaude Code in VS Code

Hosting marketing sites

Lovable

Native Swift and Kotlin apps

ModaalClaude AgentCodex CLI

Amine Benjelloun

Co-founder & CTO, Mailwarm

Thinking, research, writing

ChatGPTClaude

Development

Coding agents through Conductor

Trying different models

OpenRouter

Images and video

ExtendFrameImage and video models

Where they would start

We asked the experts what they would tell someone in their field who is starting to use AI. Their advice draws on their own fields and experience.

Elena’s example comes from her community: “Jim van der Waal, Head of Product at Polarsteps, built a personal app called MeatVault with Modaal. … He did not need an approved plan or a business case to begin, and he started with something small that he wanted for himself.”

Simon adds one more use: “Claude is great at holding you accountable. … AI can be incredible at keeping you on track if no one else is.”

Amine puts it simply:

Thank you

This is the second article in our series of expert interviews. At a time when AI is changing how we work, we think it matters to stop and reflect on what is actually changing. If you missed Part 1, six other experts in video, documentation, AI visibility, product and software answered the same questions there.

A big thank you to Simon, Amine, Rylan, Sharon, Elena and Ariadna for their answers and for sharing their experience.

A new part of AI at Work comes out every two weeks on Shift. If you’re an expert in your field and would like to take part, or you’d like to get the next parts by email, you’ll find both there.

And you: what would you never hand over to AI? Tell us on LinkedIn.

Thank you for reading all the way to the end.