What is Alma, the AI agent for Shakers talent

Four rounded blob mascots in a row on a white background: three in mint green and one in cream, with dash eyes, and a small dark-green sparkle floating between the last two
Written by
Ignacio Alonso Ferreiro Follow

Obsessed with the question "why?". That obsession led me into the world of UX, and I now enjoy knowing that my job is to make sure every decision about a digital product is aligned with what its users demand. Only then am I sure they are the right decisions. Or at least, that they are not the wrong ones.

• ...

Alma is the AI agent for Shakers talent. You find it in the side panel of the Platform or in your own assistant over MCP, and what sets it apart from any random chat is the two contexts it is connected to: Shakers', with the market a talent cannot see from the outside, and yours. You ask, it does. And when a decision is due, it brings the judgement you need. It is not a project search with a chat attached. Not a chat glued to a listing.

Alma is the AI agent Shakers builds for its talent: it carries out by chat what you would do by hand in the Platform (update your profile, see projects, prepare an application), compares your rate with that of other talent with your profile and helps you prepare interviews and client meetings. What is irreversible, you always confirm.

What Alma is, in one definition

Alma is an AI agent, and the panel is only one of the two ways to use it. The first is the permanent conversation side panel of the Shakers Platform, which follows the talent across the whole interface without taking more screen than it asks for and shrinks to a discreet bubble when it has nothing to add at that moment. On one side, the chat. On the other, the Platform. The second is MCP (Model Context Protocol), the open standard that connects AI agents to tools and data: with it, you use Alma from your own assistant, outside the Shakers website.

Inside the panel, the two pieces are the same product: what you do by chat is reflected on the Platform, and a good part of what is done by hand can already be requested by conversation, with the same validity and the same effect on the talent's account as if they had pressed every button themselves. The scope here is bigger: Alma accompanies the talent's entire relationship with Shakers, not just project search, which is the part any chat with access to a listing already solves.

Alma does not answer general questions or chat for the sake of chatting, because it is connected to two contexts, Shakers' and the user's, and that is why it can do two things a loose chat cannot: carry out real Platform actions and recommend with data the talent does not have at hand. If it only did the first, it would be a pretty shortcut to forms; the second is what justifies the whole product.

One precision about identity: Alma does not replace the talent's judgement, it equips it, and the difference matters in every decision the agent prepares, as the principles below show.

The two problems it solves

A talent's life on Shakers has two separate costs, and it is worth not confusing them. The first is operational friction: you know what you want to do, but not where to do it, and applying to a project, adding a skill or updating your rate hides no difficulty at all beyond finding where the Platform's design left it. The second is missing judgement: you do not know what you should want. Whether your rate makes sense. Whether that project fits you. What they are going to ask you in the meeting with the client.

Alma attacks both, with different tools. Friction is solved by executing. You ask for the action. It prepares it or does it. No detours. Judgement is solved by recommending with Shakers' context, which is where the market information a talent cannot have lives, because no individual professional sees hundreds of projects, rates and client decisions at once. Of the two, the second is the one that changes the experience. A shortcut to forms does not deserve half a screen; a coach with the Platform's data does.

The five principles that govern Alma

Alma's design is decided in five principles, in this order:

  • Do not disturb. Alma greets you when you arrive and, from there on, only interrupts when it has something useful to say, because an assistant that talks for no reason teaches the user to ignore it.
  • Zero loss of context. You change screen or come back hours later, and the conversation still knows what you were discussing, without starting from scratch for moving around the Platform.
  • What you do by chat, by hand too. Everything you do by chat you can do by hand in the Platform, and the list of manual tasks entering the chat keeps growing. The website stays whole.
  • Talent confirms the irreversible. Alma proposes and prepares. Applying to a project or withdrawing an application is always confirmed by you, explicitly.
  • Extensible without touching the core. Adding a new action must not demand redesigning the conversation. The list of things Alma can do grows without the agent being rewritten.

How it works inside: a catalogue of actions

Alma has a registered catalogue of actions, not a closed menu and not a scripted conversation: each action knows what it is, whether it queries, whether it writes, whether there is no way back. When you ask for something, it picks the action from the catalogue and runs it with the context of the conversation and of where you are navigating as ingredients of the decision.

Multi-step actions, such as applying to a project, are a small flow with an order: search, check the fit, prepare the application and confirm, with a trace of every step and the last step always in the talent's hands. And if an action is marked irreversible, the talent's confirmation applies by itself, without each action programming it on its own, because the protection does not depend on every developer remembering it.

Who it is for: accelerator or coach

Alma serves two profiles with opposite needs. The talent established on Shakers already knows where everything is. They want speed: apply to this project, update their availability, and no unprompted lecture from the agent, since that is what email is for. For them, Alma is an accelerator that should shrink and stay quiet the rest of the time. The talent that has just arrived does not yet know what reasonable looks like. For them, Alma is a coach supplying the judgement they lack: what rate to ask for, which project to apply to, how to prepare the meeting.

The same agent, two modes. The person draws the line, not the product.

What it can do: from reviewing your profile to preparing an interview

What Alma covers today is understood well through these four examples, although its list of capabilities keeps growing:

  • See projects. Query and filtering, with no side effects of any kind: the simple case of asking what projects are open, on what financial terms and with what deadlines, and getting the answer sorted without touching a single filter.
  • Check your rate. Query with judgement: it compares your rate with that of other talent with your profile on Shakers, something you cannot work out on your own.
  • Apply to a project. Multi-step write, with the talent's explicit confirmation and a visible effect on the Platform. The case that proves the complete mechanism.
  • Review your profile. It tells you what is missing for it to be complete and how it fits the role you are after.

And one of the most valued by talent: preparing the meeting with the client, what they will ask you, what data is worth bringing and how to tell the story of your experience. It is pure judgement. It has no manual substitute in the Platform: there is no universal guide that works for every client and every project, because preparation depends on context. For anyone preparing an interview or a pitch meeting, it is the difference between improvising and arriving with context. Shakers' AI agent does not remove the interview. It prepares it.

How it protects the talent's decisions

An agent that carries out actions on someone's behalf is judged, above all, by how it treats the ones that cannot be undone, the ones whose effect on the account, the project or the client relationship outlives the press of the button. Alma's rule is a single one with no exceptions: the irreversible is confirmed by the talent, always. There is no automatic mode for applying to a project or withdrawing an application, and that absence is a product decision the catalogue of actions itself makes impossible to bypass. Alma prepares the ground, shows what is going to happen and waits for the yes, and trust is designed like the actions: conservative by default.

How to start with Shakers and Alma

If you already work with Shakers, Alma is waiting inside the Platform. If you do not yet, the way in is the AI Builders page: it explains how projects work and what your profile needs to get started.

The shift underneath is not Shakers' alone. When execution gets cheap, what rises in value is the judgement of the person deciding, the same scissors we already explained when analysing what AI agents buy and what they do not. And knowing the other side of the project makes you a better professional: the AI Tour recap on what an AI builder is describes what that market already expects of the profiles who build with AI, the same bar a Shakers talent is measured against.

FAQ about Alma

What is Alma in Shakers?

It is the AI agent for Shakers talent: you use it from the side panel of the Platform or over MCP, it is connected to Shakers' context and to yours, it carries out Platform actions and brings judgement on rates, project fit and interview and meeting preparation.

Does Alma apply to projects on its own?

No. Alma searches, checks the fit and prepares the application, but applying to a project or withdrawing an application is always confirmed by you, explicitly. The irreversible is never executed automatically.

Do you need to know about AI to use Alma?

No. You talk in normal language and every action is reflected on the Platform, where it can also be done by hand. Chat is an alternative route, not a requirement.

Does Alma replace the Shakers Platform?

No. Everything done by chat can also be done by hand, and a good part of what is done manually can already be requested by chat. Alma adds a route, it closes none.

What does Alma have to do with interviews?

Alma helps you prepare them with the context of the project and the client: what they may ask you, what data is worth bringing and how to tell your experience.