ARQUESTRA
Why products get stuck, and who should own them.
To the Product
0:00 / 3:00
A pattern

Do you recognise this pattern?

Scroll down. It takes just three minutes.

01
You study the market.
02
You build a demo.
03
You show it to the client.
04
Everything looks good. Green light.
05
You build the MVP together.
06
Then the product gets stuck for months.
What this means for you
It is not you.

If this has happened to your team, you did not run a bad project. You ran a normal one.

The evidence
62%
of organisations are at least experimenting with AI agents.
23%
are scaling one, usually in just one or two functions.
Source: McKinsey, The State of AI in 2025, survey of 1,993 participants in 105 countries, November 2025.
Rule of thumb
Most pilots do not fail. They wait.
Where they wait

At the point of trust.

82.9%
saw AI projects delayed before production by security or compliance reviews.
81.6%
shipped AI in a reduced form because of security concerns.
Source: Enterprise Management Associates, The State of AI Friction, 2026. Survey of 150+ IT and security leaders, commissioned by Protegrity.
What this means for you
Trust is not the last gate. It is the road.

Security, compliance, reliability, contracts. Every client has its own specifications. Prove the system can be trusted and you have earned the signature. Nothing ships before that.

A review cannot pass a product nobody answers for.

The road: Ivanti, 2026

Everyone says someone owns it.

85%
say every AI agent and workflow in their IT organisation has a named, accountable owner.
42%
say that accountability is actually clear.
43
points between saying there is an owner and knowing who it is.
That gap is where the product gets stuck.
Source: Ivanti, Scaling AI in IT Operations: The Path to Maturity in 2026, survey of 1,500 IT professionals.
Rule of thumb
A name on a spreadsheet is not an owner.
The empty chair

Who owns an AI agent's access? Ask around.

28%
Security
21%
Development or engineering
19%
IT
9%
Identity and access teams
?
The chair that should be taken
Source: Cloud Security Alliance, Identity and Access Gaps in the Age of Autonomous AI, survey of 228 IT and security professionals, January 2026.
What I learned
The ones that make it have an owner.

It is not only AI agents. An API integration on a server. An internal tool. A data migration. Wherever nobody owns the last mile, the work waits.

1
76% against 46%

Where most projects have an actively engaged sponsor, 76% meet their goals. Where fewer than half do, 46%.

2
Year after year

The same report names the engaged sponsor as the top driver of projects meeting their goals. In 2018, one in four organisations named inadequate sponsor support as the primary cause of failed projects.

3
They use the word

The Standish Group calls the executive sponsor the owner of the project, and says every project needs a committed sponsor or product owner.

Sources: Project Management Institute, Pulse of the Profession, 2016 and 2018, all project types. Standish Group, CHAOS Manifesto 2013: Think Big, Act Small, IT projects.
The turn

That owner is a role, not a title.

The sponsor pays. Someone else has to carry the project every day until it ships. Product manager, project manager: the name changes. What matters is that someone takes the chair.

Whoever takes it knows this road already: the specifications that change with every client, the security review, the team that has to hold together through it.

Where does this happen?

I have seen this pattern again and again.

Working in global companies
Internal productTalksNever integrated
Working in startups
ClientsTalksToo complicated, do I really need this?
Even in my own products!
ClientsTrust and securityStuck
InvestorsProve demand first

The pattern is always the same. And it hurts, every single time.

What I would do differently
1

The right problem

Find out early whether it hurts enough. 'Do I really need this?' is a valid answer. It should arrive in week one, not month six.

2

Stakeholders

Name the owner and earn the trust of whoever signs. Trust is the road, so they walk it with you from day one.

3

The MVP loop

Build the smallest thing, show it, learn, repeat. A pilot that waits is a loop that stopped.

Who is writing

I am Pablo Ramos-Sánchez, Product Lead at Arquestra. I took it from a classroom observation to an AI agent platform on the table of a large chemical company, and made the call to keep data and compute inside the customer's environment. Before that: first hire at Exomatter, where enterprise clients adopted the product, and a PhD in chemistry from KU Leuven with Covestro.

To the Product

If you recognised the pattern, tell me where yours got stuck.

Click to copy