Dynamic pricing consulting and bespoke AI pricing systems. We audit, build and run them, from the team that authored the O’Reilly book on reinforcement learning. Start with an AI pricing readiness assessment.
The team at Winder.AI are ready to collaborate with you on your AI project. We tailor our AI solutions to meet your unique needs, allowing you to focus on achieving your strategic objectives. Fill out the form below to get started.
The willingness to confront hard problems head-on rather than optimising for superficial wins. The work was treated as true research, with careful debugging and transparent discussion of failures as well as successes. That rigour significantly increased our confidence in the conclusions. Recommendation Score: 9 / 10.
Most businesses leave money on the table because their pricing decisions are made in a spreadsheet, on instinct, or by a static rule that has not been revisited in years. Winder.AI provides dynamic pricing consulting and bespoke revenue management systems for capacity-constrained industries, from hotels to energy markets to logistics. Our team authored the O'Reilly book on reinforcement learning and partnered with Duetto, the global leader in hotel revenue strategy, to evaluate offline RL across thousands of hotels.
Your team adjusts prices a few times a week, across hundreds of products, with limited context. Every decision is a guess at the elasticity of demand. The good ones do not get celebrated. The bad ones quietly leak revenue you will never recover.
Rule-based pricing handled the world of five years ago. It does not handle the modern interaction of seasonality, channel mix, perishability, and competitor moves. The patterns that matter sit below the resolution of any rule you can write down.
You evaluated Pricefx, Vendavo, or PROS and the fit was not there. Unusual inventory, a novel channel mix, regulatory constraints, or the genuine cutting edge. You need a system built around your data, not your data forced into a vendor’s schema.
A four-phase ladder that lets you decide whether to proceed at each gate. You buy clarity first, capability second. Most engagements start with the assessment, and many stop there because that is the right answer.
Pricing AI generates the largest returns where inventory is constrained, capacity is perishable, and demand is volatile. These are the verticals we focus on, with case studies to back the claim.
Dynamic pricing for hotels, vacation rentals, and hospitality groups. Sparse data per property, strong seasonality, and a booking horizon that compresses risk into the final days. We partnered with Duetto on offline reinforcement learning across thousands of hotels. Read the Duetto case study.
Wholesale electricity, time-of-use tariffs, demand response, and EV charging. Reinforcement learning fits naturally because today’s bid affects tomorrow’s reservoir, battery state, or grid position. We applied reinforcement learning to hydro generation and market bidding for Genesis Energy.
Grocery, fresh produce, fashion end-of-season, and any catalogue where holding cost or expiry pressure forces a markdown decision. Off-the-shelf pricing software treats every SKU as if it were the same. We build models that reflect what perishability actually does to the value of a unit on day 14 versus day 7.
Two-sided marketplaces, freight, last-mile delivery, and ride-hail-like platforms. The pricing decision is sequential, the supply and demand sides interact, and a discount today changes the inventory available tomorrow. Reinforcement learning earns its place here.
Capacity-constrained, perishable, and a problem so well-studied that the published methods are now decades old. We modernise the methodology rather than replace it wholesale. Adjacent credibility: our flight scheduling RL case study on operational decision-making for a leading aerospace company.
There are three things we offer that a generalist consultancy, a strategy firm, or a software vendor cannot match. They are the reason serious teams pick us for hard pricing problems.
Phil Winder authored the O’Reilly book on Reinforcement Learning, the technique that underpins most modern pricing AI. We do not learn RL from your project. We bring thirteen years of applied experience, and our reinforcement learning consulting practice has shipped RL systems across hospitality, energy, and aerospace.
Every pricing pitch on the market promises an uplift number. We will not, until we have seen your data. Offline evaluation of pricing agents is inherently biased, and we have built the evaluation methodology that quantifies that bias rather than hides behind it. Honest answers are the entire value of an engagement at this stage.
Strategy consultancies hand you a Powerpoint. Software vendors hand you a contract. We hand you a production pricing system with the code, the evaluation framework, the runbooks, and the people who built it on the phone when something breaks. Engineering-led delivery is the only delivery model we know.
Reinforcement learning (RL) is the technique behind most modern pricing AI. It is also the technique most often misapplied. Here is when it earns its place, and when a simpler method will serve you better.
Pricing is the textbook offline RL problem. You cannot let an untrained agent set live prices to learn from the response, because the cost of a bad experiment is real revenue. Offline RL learns from your historical pricing data without ever taking a live action. The technical challenge is that historical data is biased toward the actions your previous pricing system actually took. We use Implicit Q-Learning (IQL) and related offline methods, with adaptations we developed during the Duetto engagement to handle the sparse rewards typical of booking and transaction data.
The methodology behind this work is documented in our data science consulting practice, and the supporting research drew on the wider AI consulting services we provide.
Reinforcement learning is one technique in a much larger toolbox. For many pricing problems a simpler method is faster to build, easier to defend, and almost as good. We build whichever method actually fits, including:
The honest answer is that many pricing problems are best solved by demand modelling plus a constrained optimiser, and a meaningful subset benefit from contextual bandits or causal methods. RL is the right tool for the sequential, capacity-constrained subset. We tell you which you have, then we build it.
A short, honest list. If your problem is in here, we will refer you to the right partner rather than take the engagement. This is part of the value of working with us.
We do not work on B2B negotiated pricing (Pricefx and Vendavo are mature platforms in that space). We do not write insurance primary pricing or actuarial rate filings (specialist actuarial consultancies own that territory). We do not deliver SaaS list-price strategy or commercial-model design (Simon-Kucher and OpenView are the right calls). For lending, credit pricing, and other regulated financial decisions, see our AI in finance industry hub. If you are a smaller business looking to turn a spreadsheet-based quoting process into a working engine, that is an AI agent build rather than a dynamic pricing engagement, and the agent services page is the right starting point. We do not promise specific revenue uplift numbers before we have seen your data. We do not deliver a slide deck and walk away.
If you are unsure whether your problem is in scope, the 30-minute pricing strategy call below will tell you, free of charge.
Case study
Winder.AI helped Duetto evaluate offline reinforcement learning for dynamic hotel pricing. Over five months, the engagement progressed from behavioural cloning baselines through Implicit Q-Learning experiments on real booking data, revealing where RL outperforms simpler approaches, what data quality prerequisites exist, and how to evaluate pricing agents when ground truth is unavailable.
Case study
Winder.AI helped Apartment List modernize its machine learning operations by unifying data pipelines, automating Kubeflow workflows, and introducing enterprise-grade governance. The outcome: consistent training and inference data, faster deployment cycles, and self-service capabilities that enabled Apartment List’s data science team to scale model delivery with confidence.
Case study
Using digital twin data to build flight traffic simulators and train reinforcement learning AI agents. A leading aerospace business and Winder.AI opened new horizons for dynamic, data-driven scheduling solutions that integrate with our client’s advanced flight planning technology.
We have delivered AI and reinforcement learning systems for some of the world's most recognised companies, and we bring the same engineering rigour to every pricing engagement.
AI
“AI contract review” is a popular keyword to compete for. Look, I’m doing it right now! A couple of years ago my colleagues and I half-built a contract review service prototype. We decided not to take it any further, but that was a mistake. It’s now very hot.
So hot you can easily find a wall of product pages. Sign up, upload your contracts, get results. The pitch is simple. For straightforward use cases, it works.
But what if your contracts don’t fit their templates? What if your review process has steps a product can’t model? What if your data can’t leave your infrastructure? What if your firm’s clause playbook differs from the vendor’s defaults?
This article walks through what a custom-built contract review pipeline actually involves.
AI
Legal AI adoption has accelerated. Litify’s 2025 State of AI in Legal Report found that 78% of legal professionals now use AI in some form, up from 23% just two years earlier. In Winder.AI’s 13 year history (and counting!) I have observed a similar trend first hand.
On the back of this trend, significant VC funding has attempted to capture a share of this market. $2.4 billion was invested in 2025. A tsunami of products promise to automate contract review, legal research, and document analysis. Many of them work for a while. Then firms hit the ceiling.
This article is about where that ceiling is and what lies beyond it.
AI
Not every document workflow needs AI. If you process ten invoices a week from the same supplier in the same format, manual handling is fine. But some workflows sit at the intersection of high volume, varied formats, and expensive errors. Those are the ones where AI document processing consistently outperforms human processing on speed, accuracy, and cost.
Five specific document workflows deliver a measurable advantage when you hand them to AI.
Common questions about our pricing consulting and AI revenue management services. If your question is not covered here, book a call and we will answer it directly.
Dynamic pricing consulting helps businesses build automated systems that adjust prices in response to demand, inventory, competitors, and customer context. Winder.AI delivers dynamic pricing as a complete service, from feasibility assessment to production deployment, using modern machine learning and reinforcement learning rather than off-the-shelf software.
Published industry benchmarks suggest 2 to 7% margin uplift in 12 months, equating to 200 to 350% EBIT improvement (McKinsey). Hotel revenue management vendors report +6% RevPAR in year one. We do not commit to a specific number until our pricing readiness assessment establishes feasibility on your data. Pricing AI pitches that promise an uplift figure without seeing your data are not credible.
Our AI pricing readiness assessment runs 2 to 4 weeks. A proof of concept takes 8 to 12 weeks. A production pricing system takes 4 to 9 months depending on integration complexity. We deliver in phases so you can decide whether to proceed at each gate.
Our pricing readiness assessment starts from £15,000 and typically runs £15,000 to £25,000. Proof of concept engagements run £60,000 to £180,000. Production builds start from £180,000. Ongoing managed services start from £4,000 per month plus usage-based pricing. See our pricing page for detail.
Data quality is the single biggest reason pricing projects fail. Our readiness assessment includes a data audit that quantifies sufficient price variation, transaction-level completeness, reliable join keys, and the gap between expert behaviour and customer behaviour. If the data is not ready, we tell you, and we will help scope the data engineering work first.
Reinforcement learning is the right tool when pricing decisions are sequential (today’s price affects tomorrow’s inventory) or when rewards are delayed (a discount affects lifetime value). For static-context, single-decision problems, simpler methods like contextual bandits or causal regression often work better. We pick the simplest method that fits.
If your problem fits a platform vendor, you should buy the platform. We build bespoke pricing systems for the cases where the platform does not fit, including unusual constraints, novel domains, or genuinely cutting-edge methods like offline reinforcement learning. Our build-versus-buy guide helps you decide.
Offline evaluation of pricing agents is inherently biased, because the same model that selects prices is often the one used to score them. We use a multi-metric framework covering counterfactual revenue estimation, logical constraint tests, feature importance analysis, and quantified bias adjustment. Production deployments include staged A/B or geo experiments. This is the methodology we developed with Duetto and use on every engagement.
Hospitality (hotels, vacation rentals), energy and utilities, e-commerce with perishable inventory, marketplaces and logistics, and airline and travel revenue management. We do not work on B2B negotiated pricing (Pricefx and Vendavo territory), insurance primary pricing (specialist actuarial territory), or pure SaaS list-price strategy (Simon-Kucher territory).
For regulated domains, we design systems with full audit trails, decision-level explainability, segment fairness monitoring, and override capability. We track the algorithmic-pricing regulatory landscape (FCA price-walking ban, the 2025 New York algorithmic-rent ruling, state-level US developments). For insurance primary pricing or US multi-state rent, we recommend a specialist actuarial or legal partner alongside our engagement. We have legal partners that we can recommend and work with.
The team at Winder.AI are ready to collaborate with you on your ai pricing readiness assessment project. We tailor our AI solutions to meet your unique needs, allowing you to focus on achieving your strategic objectives. Fill out the form below to get started.