product design

product design

While working in different companies and serving multiple clients I collected a lot of experience and exposure to products.

evolution of products design

When I started my enterprise career in 2013 the market looked like the golden age for product design. The companies where pivoting around products re-structuring teams around ideas, breaking hierarchies and building sub companies to be ready to treat as external entities.

company’s evaluation

For VCs investing in startups the most important moment is the company exit more than the relevance of the business or the durability of the company. The exit value is based on company evaluation which in many cases around 2020s wasn’t really tight to revenue figures.

bad return

Among all the strange evaluations we saw big investments in products who didn’t have enough market or enough maturity to disrupt. We paid with increased service costs some of these investments that didn’t bring the expected return. The question is whether the buyer or the seller had a bad understanding of the profitability of the products. To name a few:

virtual reality

hard to think you can spend hours a day with an visor on your head

5G

hard to quantify the improvement. The moment I switched it off from my self phone my battery lasted 3 times longer

web3

a lot more computation needed and unclear advantages

blockchain

I don’t see any more companies offering this service

autonomous drive

in 2015 everybody said in 3y we will stop buying private cars

SaaS

in 2026 it’s hard to see how can they be profitable

LLMs

we’ll see. It’s hard to think that a subscription model could bring long term revenues

cost saving

The easiest argument for selling to a client is cost saving.

needs

The easiest path is to start from market needs. The most important aspect is optimization and I see different areas where the industry needs a huge productivity boost:

healthcare

one of the most significant expense for the state budget

predictive

prevention is a huge cost saving which rarely get considered. Data are not collected correctly, shared, anonymized, used…

logistics

humans have many means of transportation but there is no infrastructure for goods in city centers

logistics

Logistic is a field where the optimization potential is huge but the resistance against changes is even larger.

delivery

Delivery sounds like a well established market where everything is highly optimized. I rarely order something online but when I needed some hardware components my experience was in 2025:

China

the provider didn’t have debit card nor IBAN options to pay

US

the price wasn’t considering custom fees which wasn’t clear at the time of the order

EU

the product wasn’t in stock and the price was much higher due to hypothetical custom fees

account

I had to open two different accounts for the two different regions for the same company. Each region has separated customer service, terms and conditions

delivery

I find in my mailbox a note that the courier didn’t find me when I actually was at home

tracking

I opened the different emails to find out who was the carrier but the link for the tracking ID was broken

customer service

I contacted the Indian contact center multiple times, people where kind but couldn’t help, nobody could tell me where my parcel I paid for was delivered

parcel

I went all over the shops in the neighborhood where usually parcels get delivered and finally found the product I paid for

laptop

when my company sent me the laptop they delivered it to a neighbor breaking the confidential terms of my contracts. The delivery service sent me a note instead naming a shop close by who didn’t actually received my laptop

It is pretty clear that the optimization potential is huge and the industry is not mature enough. The idea itself that running vans across the city for door-to-door activity consuming boxes and trash is not competitive against a well organized mini hub logistics is not comparable. The labor, even with automation, for door-to-door is not convenient.

agentic in software engineering

Agents can be really useful for software engineering, they are proficient in writing code and they shift the paradigm from coding to overview. This should be measurable to understand the real benefit because instead of writing the code you need you should instead:

models.json

list of models to use, their function and limitations

settings.json

describe the system you are using and the resources to allocate

system.html

how your system is configured, what servers and what volumes

system_append.html

sub-agent specific configuration

rules.html

how the agent should behave in general (test, don’t break, don’t access…)

hooks

hard coded tools to test agent behavior

agents.html

agent definition for the specific project

orchestrator

instruction for the orchestrating agent

skills.html

what each sub-agent should be capable to do

readme.html

what is the project about, scope and objectives

architecture decision records

document technical choices and evaluate alternatives

mcp

integration with other tools

tools

enable/restrict tools at sub-agent level

sessions

record history of each sub-agent run

repository

define at repository level the naming conventions and standards

PR

define a strategy for each PR review

testing

define the tests that the agents should run before committing

guardrails

how to control agent’s output

sandboxes

create sandboxes to restrict permissions allow/ask/deny

security

mitigate jailbreak and malicious prompt injection

success criteria

measurable metrics for evaluation

costs

instructions how to keep costs within a budget (the tiniest sub-agent still has a pre-push script, a fix-failing-test hook, AI code review …) trigger/scope/feedback/review

CI/CD

organize the access and CLI for each sub-agent

maintainability

check whether the code is easy to read and maintain

harness

building the system around models to turn them into work engines: tools execution, feedback, evaluation

pen-test

the engineer needs to run tests to see whether any rule breaks

Despite all of this work no agent can be accounted as responsible for anything going wrong and the engineer approving should be able to review.

IT providers

In IT you don’t have many options for infrastructure providers, in the past there were many small-size vendors of small data centers that where highly customizing solutions for their clients. Cloud won a big part of these market promising cheap and scalable solutions. As per many other business we live nowadays in an economy where we have less and less players. Investors allow few company to beat small business and they are able to rise service prices once competition is gone. It is always really important to keep the ability of switching to custom solutions in case your business will be put at risk by the increase of running costs you can’t control.

Naming few providers:

cloud

aws, azure, gcp

analytics

databricks, snowflake

LLMs

anthropic, openai, gemini

CRM/tickets

salesforce, pega, servicenow,

Risks from an external providers.

service costs

if the provider changes subscription costs your margins will instantaneously decline

lock-in

sometimes you are not allowed to create your own solution is the provider has a paid solution

quality drop

if the provider drops the quality of the service your business is directly impacted

deadlines

relying on factors you can’t control can severely affect deadlines

big size

those providers won’t adapt any solution for your case, you would need to hire externals

big catalog

the services they provide are many, it takes competences and time to understand how to configure them for your use-case.

dependencies

To design a product it is crucial to understand the chain of dependencies and compute the volatility of changes in input prices and perform some stress tests.

drafts

We are working at different product drafts:

barrique app

social entrepreneurship