AKA Studio Alchemy
Studio doesn't just help you develop products. It develops them with you.
The biggest change to AKA's AI platform for food R&D since launch. The last generation informed the work: it found the data, compared candidates and drew the charts. Alchemy does the work: it creates projects, fills briefs, generates batches of candidate recipes and writes the reports, inside the limits your technologists set.
Four big changes
Alchemy turns AKA Studio from food product development software that answers questions into an AI that moves the work forward. The recipes it produces are drafts, and your expert still decides: it augments the technologist, never replaces them.
It does the work
The assistant is agentic: it creates prototypes, builds batches, updates the brief and reads a product label from a photo, always asking before it acts. AI formulation that acts, not just answers.
Development runs in batches
Product development moves through purposeful batches instead of a straight line. Each batch is measured at a round table, and sensory data feeds straight into the next iteration.
You hold the levers
Layered constraints with hard and soft toggles, enforced numeric limits, and full transparency: every candidate shows its sources and an honest compliance checklist, with numbers.
Value from day zero
No data onboarding required to start. Ideate, formulate and report from your first session; connect your own R&D data when you're ready and everything gets sharper.
The loop, finally visible
Before, you uploaded your data and asked questions. Now the whole development cycle lives on one screen: ideate, formulate, run the sensory test, audit for regulation, and go again. That loop is how food R&D actually works, and Alchemy is built around it.
Ideation
Explore directions beyond your current ingredients and processes, check the market angle with live research, and turn a concept into a project with the brief pre-filled.
Formulation
Generate batches of candidate recipes against your brief and constraints. Hard limits are held and flagged if they can't be met, and you steer every candidate.
Sensory
Round tables capture tasting feedback live, batch by batch, and the insights feed the next batch with no downtime.
Audit
A regulation-focused mode drills into the detail, checks candidates against constraints, and generates QA/HACCP and project reports on demand.
All 63 improvements, listed
Every shipped change in this generation, area by area. No highlights reel, the whole list.
Chat & Assistant
- Redesigned chat with a cleaner input and a dedicated context area
- Queue messages while the assistant is still answering
- Agentic assistant that performs platform actions itself: create, duplicate a prototype, create a batch, change the brief and more
- Always asks before it acts
- Smart suggestion pills: context-aware quick actions as you chat (generate a report, summarize, create a recipe from this)
- Dynamic quick hints that change with the active context and scope
- Resizable chat panel
- Keyboard shortcuts with a quick-reference cheatsheet and hover hints
Working modes
- The assistant auto-switches into a mode based on the problem: Formulate, Ideate or Audit-for-regulation
- Each mode changes the prompts and model temperature: the auditor drills into detail, ideation runs hotter
- Context handover between modes: focus and context carry over when the stance switches
Context & search
- Control the context scope (project, platform or the external web), shown like breadcrumbs
- Online search toggle, activated per entry point
Attachments & label reading
- Attach images and documents directly in chat: images go to your media library, documents to Knowledge
- Label reading: snap a photo of a product label and the assistant reads its ingredients, nutrition and claims, saved, tagged and searchable
Charts & visualization
- Charts in chat: ask for a plot in plain language and a real chart renders in the conversation
- Full range of chart types, including radar charts for comparing prototypes across attributes
- Fast baseline charts for common requests, complex plots routed to the agent
Reports
- Generate reports in plain language ("make me a cost report") as a downloadable PDF in chat
- Charts can be included inside exported PDF reports
- Built-in report types: project summary, QA/HACCP, protein quality
Recipe generation
- Suggest or modify recipes with a formulation-strategy planner and full sources behind every suggestion
- Control the wizard's context scope: project, platform or external web
- Uses your curated candidate sets as starting points
- Interactive suggested-recipe cards: hover to add to a project or copy
- Add to an existing project or create a new one right in the conversation
- Generate a single recipe with no batch (a starter path), push it to the project or export to CSV
- Generate multiple candidates at once and manage them as batches
Batch generation: explore/exploit
- A card showing exactly what will run (candidates, anchor, objectives, constraints), editable in place
- Choose the number of candidates
- Explore/exploit control: constrained vs exploratory candidates, soft constraints become guidance
- Steer individual exploratory candidates wherever you want them to go
- Anchor selection: a prototype, the project, general, or your choice
- Candidates built in parallel
- Live run tracker with real status per candidate and learning receipts
- Source transparency: each candidate shows what it drew from (knowledge library vs general food science)
- An honest compliance checklist per candidate: what held vs where it deviated, with numbers
- Land results directly as a named batch, with an optional round table
Batches
- Development moves through batches rather than linearly; each batch has a purpose and is measured at a round table
- Arrange and group prototypes into batches from the Batches tab
- Drag-and-drop prototypes into batches; create new batches easily
- Reassign a prototype to a different batch
- Generate a batch directly from an existing prototype: it prepares an optimized recipe and jumps ahead
- Status indicators per batch, updating live
- Per-batch AI summary: the AI reads the prototypes and learns the target and best-fit candidate
Round tables
- Create round tables straight from the Batches tab, bringing the batch's prototypes with them
- Launch a round table from inside a batch; auto-picks the relevant batch and prototypes
- Run multiple round tables to compare, with summaries collected batch-wise
Brief & constraints
- Layered constraint system (base, project and chat-level) merged with clear provenance
- Dedicated Project Constraints section in the Brief with editable priorities and inline editing
- Constraints auto-derived from brief changes; the chat detects new or reduced constraints on the fly
- Hard vs soft toggle: hard constraints the AI won't break, soft are best effort
- Nutrient ranges keep distinct minimum and maximum values
- Hard numeric limits enforced: "under 400 kcal" is calculated and held, and the assistant tells you if it can't be met
- Nutritional targets moved into the Brief
- Reads brief documents in Word and Excel
Knowledge, discovery & navigation
- Three-layer knowledge system: AKA knowledge, organization knowledge and project knowledge
- Home-page assistant for discovery: give ideas, refine directions, check the market angle
- Create a project directly from a discovered concept, with the brief and description pre-filled
- Reworked navigation panel with dynamic sections
- The plan/roadmap view is now optional
- Multi-tab navigation within a project
- Cleaned-up prototype page with more room to run the formula
Still to come in this generation: Skills, DOE-driven explore/exploit, and in-assistant label generation.
What is AKA Studio Alchemy?
AKA Studio Alchemy is the third generation of AKA Studio, the secure AI platform for food R&D from AKA Foods. It combines AI formulation, sensory data capture and regulatory auditing in one food product development platform, so food, beverage and ingredient companies develop better products faster and at lower cost.
Unlike generic AI tools, Alchemy works inside the limits your food technologists set, shows the sources behind every suggestion, and is validated by experienced food technologists in AKA's own food, sensory and analytical labs. Client data stays private, siloed and SOC 2 protected, and is never used to train outside AI models.
See Alchemy on a real brief: yours
Value from day zero, no data onboarding required.