What holds up today and what doesn't
The applications that work in a salon share one characteristic: they handle a narrow, verifiable task. Reading proportions on a front-facing photo, separating hair from the background to read the colour, structuring a sequence of questions. In all of those the error is visible immediately and correctable.
The ones that don't hold up are those requiring judgement on information the system doesn't have: assessing whether hair can take bleaching, understanding whether a client is asking for change or reassurance, knowing when to say no. Those are the points where a wrong suggestion costs more than the automation is worth.
The practical criterion is therefore simple: introduce AI where the error shows and can be fixed in seconds, and keep it out of the points where a wrong proposal would be discovered once the service has started.
Where AI holds up in daily work
The chart is qualitative and shows how far each application is reliable enough to be used in front of a client.
What to introduce first
The table ranks the applications by the ratio between immediate benefit and adoption effort.
| Application | Benefit | Adoption effort |
|---|---|---|
| Consultation structure | Makes the team's method uniform | Low, it changes the order of questions |
| Face reading | Makes the proposal explainable | Medium, requires correct capture |
| Written look plan | Makes the proposal verifiable | Low, it's a document |
| Automatic service suggestion | Unreliable | High, needs correcting in front of the client |
How to introduce it
One application at a time
For at least a month before adding another.
Start with the simplest
Consultation structure, not image generation.
Check before showing
Five seconds on the output.
Measure the change
If nothing changes in the work, remove it.
How to introduce it
- One application at a time, for at least a month.
- Start with the one with the lowest adoption effort.
- Check the output before showing it to the client.
Introducing several things at once is the mistake that sinks adoption. The team can't tell what's working and what isn't, and at the first busy week they go back to the previous method, which is faster because it's already automatic.
The rule of checking before showing applies to any generated output. Five seconds spent on the outline of a look plan or on the coherence of a reading avoids the only truly costly error: putting in front of a client something she recognises as wrong.
From discourse to guided consultation
Saloria uses AI as consultation infrastructure, keeping the stylist and the relationship at the centre. It doesn't replace management software, doesn't promise realtime AR and doesn't turn look plan into certainty. It brings method to the moment when client and professional decide the look together.
Frequently asked questions
Where should you start?
With the structure of the consultation: it has the lowest adoption effort and the most immediate effect on team consistency.
Can AI suggest the service to propose?
Not reliably: it doesn't know the real condition of the hair or the client's context. A suggestion has to be checked anyway, and correcting it in front of the client costs more than it saves.
How long before you see an effect?
About a month of consistent use on a single application. Introducing more than one makes it impossible to tell what's working.