We’re in a cycle of continuous reformation.
Supply chain disruptions mean last year’s key ingredients are unavailable or harder to get today. Raw material costs rise faster than retailers will accept price increases. A regulatory change - whether it’s allergen thresholds, rounding rules, or nutritional claim substantiation - means thousands of products must be reconsidered. Consumer preferences shift. And the pressure to innovate, to reformulate toward health or sustainability claims, never stops.
For most manufacturers, reformulation is painful because the foundational infrastructure for managing it (e.g. spreadsheets, email threads, and institutional knowledge held by formulators who may or may not stay in their roles) was never designed for this workload.
Reformulation today is painfully linear.
Andy Stark, GM of R&D workbench at Specright
A medium-sized food company managing 5,000 SKUs with 30 ingredients per formula carries roughly 150,000 ingredient-formula relationships. Trying to manage these accurately, in a spreadsheet-based system, is architecturally impossible.
Where reformulation breaks today
The pain points cluster around three areas:
1. The data integrity gap
A single ingredient change, be it an allergen declaration update, a revised nutrition profile, a new supplier specification, or something else, must theoretically propagate to every formula it touches.
In practice - across a network of disconnected files -it rarely does.
The result is a silent compliance liability that surfaces during audits, at retail, or worse - leading to costly product recalls, retailer delists, or significant regulatory fines.
2. The institutional knowledge gap
Why was an ingredient originally chosen? What was tried and abandoned? What did a failed trial teach the team? None of this lives in a system; it lives in emails and in people’s heads.
When a formulator leaves, so does their context.
Each reformulation cycle starts from partial information, and past decisions are reinvented rather than learned from.
3. The sequence problem
Reformulation today is painfully linear. A formulator drafts a concept, often manually checking feasibility. That gets handed to the regulatory team to validate claims. Marketing weighs in on positioning. Procurement checks cost. Quality reviews potential risks.
By the time insights from regulatory or procurement loop back, the formulator has already moved on to the next project. Rework cascades.
Where AI changes the reformulation equation
AI doesn’t solve all of these problems at once. But it transforms the parts of reformulation where speed and pattern recognition matter most.
Ideation and ingredient selection
This is where AI adds the most immediate value. Rather than starting with a blank spreadsheet or a historical reference, a formulator can describe a product goal in natural language: ‘I need a plant-based meat analogue with 15g protein per serving, clean label, and shelf-stable for 18 months.’
AI can instantly:
- Suggest ingredient combinations that achieve the nutritional target
- Flag which combinations work within the category constraints you define
- Cross-reference your actual ingredient library, so suggestions aren’t theoretical, they’re grounded in ingredients you already use and that are already priced and sourced
- Show why it made those suggestions, so the formulator retains control and critical thinking
This collapses weeks of exploratory research into days. But more importantly, it reframes AI’s role - not as a black box, but as a research assistant that shows its work.
Compliance and labelling validation
Regulatory teams spend an enormous amount of time answering the same questions repeatedly:
- Does this health claim meet EFSA substantiation requirements or UK legislation (Food (Standards) Regulations 2011)?
- What nutrient declaration and rounding rules apply under UK FIC regulations vs. EU Regulation 1169/2011?
- Is this allergen declaration (Big 9 + sesame) consistent across suppliers and compliant with labelling requirements?
- Are ingredient names correctly aligned with permitted EU/UK nomenclature?
- Do our Nutrition Information Box formatting and decimal places meet UK Food Standards Authority guidance?
Historically, these checks happened after formulation was largely locked in. Now they can happen during development, in real time.
If a formulator makes a change that inadvertently violates a claim substantiation rule, allergen declaration, or nutrient rounding requirement, they know immediately. Rework is caught early, not discovered weeks later at regulatory review or during an FSA inspection.
The speed improvement is significant, but the risk reduction is more important: compliance failures become observable before they go to market, avoiding costly recalls, retailer delists, or regulatory enforcement action.
Optimisation and substitution
Once a formula is drafted, AI can recommend ingredient substitutions that maintain nutrition and performance while addressing constraints - whether those are cost, allergen status, or sustainability metrics.
Unlike generic suggestion engines, these recommendations are grounded in:
- Your actual ingredient specifications (not average data)
- Your supplier relationships and plant approvals
- Your historical formula database (so suggestions learn from what worked in similar products)
- Regulatory constraints specific to your markets
A formulator can iterate through multiple scenarios in an afternoon rather than over the course of weeks of sourcing, sampling, and testing.
How AI changes the sequence of reformulation
AI-augmented approach: Concurrent insight delivery.
With AI integrated into the formulation workflow, regulatory context, cost insights, and risk flagging are visible to the formulator as they build.
This doesn’t eliminate review steps, but it transforms them from gatekeeping to optimisation. Regulatory isn’t asking, ‘Is this compliant?’ (it already is). They’re asking, ‘How can we position this more strongly?’ Procurement isn’t asking, ‘Can we source this?’ They’re offering real-time cost comparisons while the formulation is still plastic.
The net effect is a collapse of the linear sequence into something closer to parallelisation. Formulation and validation aren’t sequential; they’re concurrent.
Which reformulation decisions see the biggest wins
Our observation from working with food manufacturers across categories show:
Ingredient swaps: AI wins decisively. Cross-referencing cost, allergen status, nutrition contribution, and supplier availability is laborious for humans and effortless for AI.
A formulator that might have considered two or three alternatives now considers eight or ten in the same time window, raising the floor of the ‘best available option.’
Estimated speed improvement: 60–80%.
Nutritional optimisation: High value, but not as dramatic. AI can model how micronutrient changes ripple through a formula, but ultimately a formulator still needs to validate taste, texture, and processing behaviour in the lab. AI reduces analytical workload significantly, but doesn’t eliminate lab iteration.
Estimated speed improvement: 40–50%.
Cost reduction: Moderate impact. AI can identify cheaper ingredient alternatives that maintain the nutrition profile, but procurement and supplier qualification still take time. What AI does well is surfacing the opportunity earlier, so the sourcing and testing happens in parallel with product development rather than as a last-minute scramble.
Estimated speed improvement: 30–40%.
Compliance and claims: Highest impact on risk reduction, good impact on speed. This is where AI’s rules-based compliance engine (as opposed to probabilistic LLM guessing) matters most. A claims matrix that human regulatory teams might review in three to four days AI can validate in seconds, with zero hallucination.
Estimated speed improvement: 70–90%.
How AI reduces reformulation risk
Risk in reformulation clusters around three categories:
Compliance risk: The biggest is silent non-compliance; a formula that appears to comply but doesn’t, discovered after launch.
AI doesn’t eliminate regulatory review, but it shifts the discovery point earlier. If a claim is unsustainable, a rounding error exists, or an allergen declaration is incomplete, the formulator sees it during development, not after market launch.
Data integrity risk: AI-powered systems with a deterministic logic engine for calculations eliminate propagation errors. When an ingredient changes, dependent formulas update automatically rather than requiring manual re-entry across 50 spreadsheet files. Compliance risk from outdated data collapses.
Innovation risk: Formulators often stick with ingredients and combinations they know, partly because exploring alternatives is time-intensive. AI reduces the friction of exploration. Teams innovate more- and more conservatively, because they can validate ideas without committing resources to sampling and testing first.
The compounding effect: AI systems create audit trails. Every decision - every substitution, every version - is logged with rationale. When a regulatory audit happens, the documentation is complete and time-stamped, not reconstructed from email threads years later.
Case study: from months to weeks
Consider a real scenario we observed with a major European CPG manufacturer reformulating a cereal bar for the UK and EU market in response to rising commodity costs and stricter sugar targets (aligning with the Soft Drinks Industry Levy and emerging category standards).
Without AI approach (typical timeline)
Week 1–2: Formulator explores alternatives manually. Sources three or four comparable sweeteners (accounting for EU authorisation status, UK permitted ingredients). Gets pricing and supply commitments.
Week 3–4: Samples arrive. Initial taste/texture testing. Formulator adjusts ratios. Iterates manually through nutrition calcs in Excel.
Week 5–6: Initial results go to quality and regulatory for review. Quality flags potential processing issues. Regulatory notes concerns about health claim substantiation under EFSA rules.
Week 7–8: Rework. Formulator adjusts. Back to quality. Back to regulatory. Regulatory confirms UK FIC compliance and allergen declarations.
Week 9–12: Procurement weighs in with UK and EU supplier options. Sustainability team reviews sourcing. Labeling team designs multi-market variants (UK, Ireland, EU). Another round of reformulation due to conflicting nutrient declaration rules.
Total: 12 weeks before you’re ready to commit to a test batch. Risk of non-compliance with UK Food Standards Authority or EFSA throughout.
With AI-augmented approach
Day 1–2: Formulator inputs the brief into the workbench, specifying target sugar reduction and market scope (UK, Ireland, EU). AI suggests seven alternative sweetener blends that hit nutrition targets, cost constraints, and EFSA/UK compliance criteria - all from ingredients already in the library and pre-screened for permitted status. Formulator narrows to three best candidates.
Day 3: Formulator builds three formula variants. Real-time nutrition calculation shows implications for macros, micros, and declared claims. EFSA substantiation rules and UK nutrient rounding rules are flagged automatically. Formulator refines based on that feedback before regulatory ever sees it.
Day 4: Regulatory review. No surprises. Health claims are already substantiated. Allergen declarations are complete. FIC compliance is confirmed. Formulation is locked.
Day 5: Procurement runs cost scenarios across UK, Irish, and EU suppliers. Plant-specific sourcing alternatives are already modeled in the system, so procurement can see real-world cost variance and lead times. Recommends the optimal variant.
Day 6: Quality reviews. Processing parameters are already linked to the formula. Manufacturing instructions are auto-generated for UK and EU facilities. No rework needed.
Total: Six days to a locked, validated, cost-optimised, multi-market formula ready for test batch. Compliance risk drops dramatically. Sampling cost drops because you’re testing fewer candidates more confidently.
The time savings here isn’t just about speed. It’s about collapsing rework cycles. Because EFSA substantiation, FIC compliance, and cost are visible during formulation, not after, the number of iteration loops needed drops from six or seven down to one or two.
The same manufacturer that might develop 40 new SKUs per year can now develop 60 with the same team, because each project absorbs less calendar time and fewer person-hours on non-lab work.
Where AI still has limits
Here’s what AI can’t do, and where skilled formulators remain indispensable.
Taste, texture, and sensory development
AI can model how ingredient changes affect nutrition and cost. It cannot taste. A reformulation that hits nutritional targets and costs 10% less may be organoleptically inferior. Sensory evaluation and consumer testing still require human judgment and real product samples.
Novel processing and scale-up
AI can flag that an ingredient substitution might affect mouthfeel or pourability. It cannot fully predict how a formula will behave in a commercial-scale dryer, extruder, or high-shear mixer. Process engineering and scale-up testing remain essential.
Regulatory edge cases
Most regulatory decisions are predictable. But FDA or Health Canada sometimes issue new guidance in a grey area. AI trained on existing regulatory frameworks can’t divine future ones. That requires a regulatory expert with intuition about how agencies think.
Innovation judgment
AI can suggest ingredients based on patterns in your database. It cannot decide whether a trend is a lasting consumer preference or a flash. That strategic judgment - deciding which reformulations are worth pursuing, and why - comes from a formulator who understands your brand, your customers, and your business.
Institutional knowledge and rationale
The most valuable aspect of a formulation is not the final numbers; it’s the why behind them. Why did you choose this thickener over three others? What did earlier trials teach you about binding in high-fibre blends?
That knowledge lives in conversations, experimentation, and critical thinking. AI is a tool that lets your best formulators spend more time on these high-value decisions and less time on algorithmic grunt work.
The human-AI partnership
The future isn’t AI replacing formulators. It’s formulators using AI to move higher—away from spreadsheet management and manual calculation, and toward strategy, innovation, and the judgment calls that require taste, intuition, and accountability.
The implications for your R&D team
If you’re reformulating today, the competitive advantage isn’t in being faster at the things you already do. It’s collapsing the sequence of development so that formulation, validation, and optimisation happen in parallel rather than in series.
That requires a system that connects formulas, ingredients, and regulatory context in real time. It requires calculations you can trust. And it requires making the formulator’s work visible to procurement, quality, and regulatory teams so they can contribute insights while there’s still time to act on them.
The manufacturers leading reformulation cycles now aren’t reworking less. They’re reworking earlier, faster, and with full visibility into trade-offs between cost, compliance, and innovation.
AI makes that possible. But only if the foundation - the ingredient data, the formula structure, the regulatory rules - is already accurate and connected. Spec-first systems make that foundation possible.
About the author
Andy Stark, is the general manager of R&D workbench at Specright.