Straight answers, before the first call.

Start with the free AI Readiness Assessment. It takes about five minutes, scores your organization across the dimensions that decide whether a data or AI build succeeds, and returns a prioritized list of what to fix first. If you would rather talk it through, a first conversation with a senior advisor is about the work, not a pitch.

In your cloud, on your accounts, with your data staying in your environment, and what we build is yours to run. We are model-neutral by policy and hold no exclusive vendor incentives, so the architecture recommendation follows your data sensitivity, volume and existing footprint rather than a partner quota. Where your team needs more depth to keep pace, we extend it with senior specialists rather than replacing anyone.

Fixed scope, fixed fee, and a defined decision at the end. Most clients begin with an Assessment Sprint, an AI & Agent Pilot or a Data Platform Accelerator depending on where their uncertainty sits, and most start within two to three weeks of a signed scope. Larger platform engagements are scoped individually, but we would normally run a bounded program first so the scope rests on evidence rather than assumption.

Data Products LLC is an AI and data consultancy headquartered in Chicago, working with clients across the US and internationally. We help leaders decide which data and AI investments are worth making, then the same senior team builds them. More about us.

Data is any information an organization tracks or stores, numeric or not: transactions, records, documents, images, sensor and building readings, finance, time sheets, customer orders and more. If a decision depends on information you hold but cannot easily use, it is worth a conversation. The free assessments are a quick way to find out where you stand.

Eight practices: Strategy & Governance, Data Engineering & Platforms, AI & Machine Learning, Generative AI & AI Agents, BI & Data Visualization, AI Security & Cyber Risk, AI & Data Literacy, and Staff Augmentation & Expert Services, plus six fixed-scope starter services. See all practices.

Yes. Our AI & Data Literacy practice builds role-based programs for executives and teams, measured by how the skills are used afterwards rather than attendance.

Use the contact page to book a thirty-minute scoping call or send a message. You can also email sales@dataproducts.io.

See articles, Executive Bytes, the Data Simplified podcast and our case studies.

The study of structured, unstructured and complex data using algorithms, statistics and mathematical tools to extract useful insight.

Building systems that learn from data and improve over time without being explicitly programmed for every case.

Building systems that perform tasks that normally need human intelligence, such as perception, language, reasoning and decision-making.

The plan for how an organization will use its data and data capabilities, aligned to its business priorities and technology.

The methods, processes and tools that turn raw data into insight for business decisions, getting the right information to the right people at the right time. More definitions are in the data glossary.

Yes, and there is no email gate. You see your complete result on screen — every dimension score, the strengths, the gaps and the priorities — without giving us an address. At the end you can ask us to email you a PDF of the report, which is an offer rather than a condition.

Scoring happens in your browser and nothing leaves it unless you ask for the report by email. If you do ask, your address and your dimension scores are sent so we can send the report. Your individual answers are never transmitted.

AI Readiness is eighteen questions and takes about seven minutes. AI and Data Literacy runs two tracks in one sitting, about twelve minutes. Data Maturity and the Cyber Risk Screen are fifteen questions each and take about six. You can move backwards and change an answer at any point before you finish.

Take the one matching the question you are actually being asked. If someone wants to know whether an AI programme will survive contact with production, take AI Readiness. If adoption is the problem rather than tooling, take AI and Data Literacy. If the doubt is about the data underneath, take Data Maturity. If the question came from security or audit, start with the cyber screen.

A questionnaire scores what you believe about your organization. That is genuinely useful — it shows which dimension is weakest and gives you language for the budget conversation — but it cannot inspect your data, your contracts, your architecture or your actual run costs. Verifying those is what a paid diagnostic does.

No. There is no newsletter subscription attached to any assessment, and no tracking pixel in the emailed report. If you ask for the report, you get the report.

From first principles: the baseline today, the intervention, its total cost, the adoption and process change it needs, the measurable outcome, and the value realized. Total cost covers data remediation, integration, workflow redesign, infrastructure, change management, oversight and run cost, not only the model. Return can be revenue, margin, cycle time, capacity, customer outcomes, risk reduction or avoided cost. We keep the business case and value realization separate: one is what the investment should return, the other is whether it did.

First each idea is articulated well enough to compare: the business problem, the user and process, the outcome and its baseline, the data required, an owner, the dependencies, the risk and a success measure. Then the portfolio is ranked on value, strategic fit, readiness, time to value, dependencies, cost, adoption requirement and risk. The result is a funding decision, not a scored spreadsheet.

Business outcomes tied to the use cases that move them, the capability gaps in the way, the data and platform dependencies, the operating-model changes, the governance each stage needs, the funding gates and the delivery sequence. It is what connects strategy to the engineering and AI work that follows.

Against your systems and the people who run them, with sources cited, across data, architecture, technology, talent, process, governance, operating model, adoption and value management. The free online assessments are a self-diagnostic for orientation. An engagement assessment replaces your answers with evidence, and exists to answer one question: what can we realistically do next?

By producing the evidence oversight relies on: what leadership knew, which alternatives and risks were considered, who owned the outcome, what supported the decision, and what would cause it to be revisited. The decision log is the mechanism. We do not offer legal advice; we build the decision structure and the record.

A single heavy review for everything does. Tiered review does the opposite: the low-risk majority takes a fast lane with a short record, and depth goes where a use case touches sensitive data, affects people or acts on its own.

By making it concrete. Can the data be relied on? Can a decision be reproduced and explained? Is the source of the evidence known? Do people know when a human must step in? Is there a trigger for re-review when performance changes? Each becomes a control with an owner.

No. Most policies have the right intent and no decisions in them. We keep the intent and add the criteria, thresholds and named owners, then re-run past requests through it so you see it decide rather than defer.

NIST AI RMF for structure, ISO/IEC 42001 for the management-system shape auditors recognize, and the obligations that actually apply to you: federal and state requirements, sector rules and your contracts. Your own policies are the starting point.

The diagnostic: every component through the seven tests, with its keep, fix or build call and the reason. A working foundation on the chosen platform, with its access and governance model. The first pipelines live, built as the pattern for the rest. And the runbook, walked through with the team that will own it. Two to four weeks, fixed scope, with a fixed fee agreed in writing before work starts.

Usually not. The warehouse is often sound; what is broken is around it, such as ingest that fails silently, missing lineage, or definitions that drifted apart. Every component goes through the same seven tests. What passes is kept, what is worth saving is fixed, and only what does not exist is built.

No. Migrations usually stall on things nobody designed: a pipeline pattern, an access model, one set of conventions across two teams. We find the missing pattern, build it once, and what is stuck follows it. What already moved is kept.

If you have already chosen, we build on it. If you have not, the decision is made on the diagnostic: your existing cloud footprint, the workloads the platform has to carry, and the team that will operate it. We work across all three, and the recommendation is not conditioned on an exclusive vendor relationship.

That is the seventh test and the last acceptance criterion. Your team gets the runbook to operate, extend and add a source, is walked through it on the platform itself, and adds a new source from the template before handover.

No. The accelerator proves the pattern on a bounded scope. The practice exists for what follows: full platform builds, migrations of every source and workload, integration and master data, and the DataOps that keeps a platform reliable once it is live.

Often the constraint is one layer, not the whole estate: the storage and compute that documents, retrieval and models need. We modernize that layer and carry forward the semantic models, history and access rules that still work.

Unit economics is one of the seven tests. Budgets, per-run cost tags and alerts are built in from the start, so the cost per pipeline run and per query is known at the volume you are planning, before the bill arrives.

The whole system: data pipelines, features, training and serving, monitoring, the integration into the workflow, and the handover that lets your team run it. A model on its own does not change a decision.

By validating against the decision, not the dataset. That means time-based splits, backtesting, leakage checks, segment-level performance, and a threshold set with the people who act on it. Monitoring and drift detection are in place from the first release.

If the answer is a number, a score, a ranking or a plan, it is usually machine learning. If the work is reading, writing or conversing over documents and knowledge, it is usually generative AI. Many systems use both, and we scope them together when they do.

By looking at it, early. Data readiness is the first technical step: access, quality, definitions, and whether the outcome you want to predict has been recorded reliably. If it has not, we say so before modelling starts, and the Assessment Sprint exists for exactly that question.

At the workflow, against a baseline set when the decision is framed: cycle time, throughput, loss avoided, false positives, conversion or cost per outcome. The measure is agreed before engineering begins, so the result is not argued about afterwards.

Yes. Access controls, audit logging, model documentation, explanations and review routines are scaled to how much the decision matters. We build the controls and the record; we do not offer legal advice.

It is monitored for performance, drift and data quality, with a named owner for alerts and an agreed trigger for retraining. Your team runs it, with our support for as long as that is useful.

A running GenAI system on your data, handed over with its source. The evaluation harness: your test set, scoring logic and acceptance thresholds. Measured unit economics at expected volume. And a scale, revise or stop decision presented with the evidence behind it. Two to four weeks, fixed scope, with a fixed fee agreed in writing before work starts.

No. The pilot is where most clients start because it produces evidence quickly, but the practice exists to put systems into production and keep them there. That covers production builds, integration with your identity, data and review path, and AI Change Assurance for agents that are already live.

We build the evaluation harness before the system. Three layers: a labeled test set with ground-truth answers, model-graded scoring for open-ended quality, and human review for edge cases. Every prompt, model or retrieval change re-runs it.

It depends on data sensitivity, cost per call at your volume, and your existing cloud footprint. Regulated data usually points to a model hosted inside your own cloud tenancy, under enterprise terms. High-volume simple tasks often run cheaper on small open-source models. Complex reasoning justifies a frontier model. Most systems we ship route across two or three. We hold no exclusive incentives; the recommendation follows the workload.

Hallucinations: grounding through retrieval, structured output constraints, and confidence-thresholded refusal. PII: classifier-based scrubbing on input and output. Prompt injection: input sanitization, channel separation between system and user messages, and output validation. We red-team every system before launch.

Model change is routine, not exceptional. One major cloud platform documents deprecation at twelve months and retirement at eighteen, with standard deployments auto-upgraded region by region and a 410 Gone response at retirement. AI Change Assurance builds a replay harness so you can prove a new model behaves acceptably against recorded production traffic before it reaches customers.

There is currently no broad regulatory requirement that enterprises revalidate a generative or agentic application every time an underlying model changes; certain regulated systems carry change-management or reassessment obligations, but those are use-case and jurisdiction specific. The stronger argument is commercial: an undetected change in agent behavior affects customers, revenue and operational cost well before it attracts a regulator.

Four reasons, consistently. The pilot ran on clean sample data and real data is messy. There was no evaluation harness, so nobody could prove the system was good enough to ship. Governance was deferred and legal blocked launch. And cost per transaction was never modeled, so finance stopped it at scale. Every engagement in this practice is built to close all four.

A RAG system retrieves from your data and generates a grounded answer: one turn, one answer. A copilot lives inside a workflow and helps a user complete tasks: many turns, one user. An agent takes a goal and acts: planning, calling tools, retrying, escalating. Most enterprise builds combine two of the three. The right pattern depends on the task and how much autonomy your risk posture allows.

Usually the one you already run. We build in Power BI, Tableau and Looker, and in embedded analytics platforms. We do not resell licences, so the recommendation is not tied to a product.

Usually not. Many reporting problems are definition and ownership problems, and can be fixed on the platform you have. Where the data underneath is not sound, we say so early, and the Data Platform Accelerator exists for that case.

By building it for a decision someone already makes, with that person in the room, and by retiring the report it replaces. After launch, where usage data is available, we track who opens it and whether the meetings it was built for rely on it.

Yes, when people explore governed measures rather than raw tables. The semantic layer does the hard part once, so two people asking the same question get the same answer.

That works well on top of governed definitions and badly without them. The definitions are what stop a natural-language tool from answering confidently and wrongly. We build the semantic layer first and add plain-language questions where they earn their place.

Usually fewer than you have now. A modernization consolidates overlapping reports into a smaller set, each with an owner and a purpose.

No. Your program stays as it is. We add what it was never asked to cover, and hand each finding to the owner who would normally close it.

A penetration test looks for ways into a system. AI red-teaming looks for ways to make a model misbehave once you are legitimately talking to it: leaking data, ignoring its instructions, or taking an action it should not. You usually need both.

Usually the opposite. Most organizations are running more AI than their security team has reviewed, because it arrives inside tools people already had. The inventory is often the most useful finding.

NIST AI RMF, the NIST Cybersecurity Framework and ISO/IEC 27001 for governance, the OWASP GenAI Security Project and MITRE ATLAS for testing, SOC 2 readiness where customers require it, and your own policies as the starting point. Sector obligations are mapped in where they apply to you.

No. We do not run a security operations center or a managed detection service, and we are not a general penetration-testing firm. We secure AI and the data around it, and work alongside the teams and firms that do the rest.

No. We build the controls and assemble the evidence an auditor will ask for. Certification is done by an accredited body, and we do not offer legal advice.

As a new kind of user. Each one gets an owner, the narrowest access that lets it do its job, credentials that expire, and a log of what it did. Irreversible or high-consequence actions require a human approval, at a threshold you set.

Data literacy is reading, questioning and acting on numbers. AI literacy is knowing what an AI system can and cannot do, using it safely and checking what it returns. A team can be strong in one and weak in the other, which is why we score them separately.

By what people do afterwards. We take a baseline before the program, measure again after, and look at the use, quality and judgment the program was for. Attendance and completion are recorded but are not the result.

Usually the executive team. People adopt what their leaders visibly understand and ask about, and the leadership session decides what the rest of the program should prioritize.

No. We design and run programs on NuScienta, on your own learning system, or on both. NuScienta can also be used on its own, without a consulting engagement.

No program does that on its own. We build role-appropriate literacy measures and keep the evidence of what was done, which is what an organization can show. Whether that meets a particular obligation is a legal question, and we do not give legal advice. Our guides to the EU AI Act and Illinois AI law set out what each covers.

Both. Executive sessions are usually in the room. Workforce tracks combine instructor-led sessions with online courses, so people practice between sessions on their own work.

Yes, and it should. Exercises are built on the systems your people already use, inside your policies, so the skill transfers to the job on the first day back.

The bench is nearshore, largely in Latin America, working in time zones that overlap the US working day. Engagements are run and overseen from our Chicago headquarters.

Three things before anyone is put forward: a technical screen run by a senior practitioner, a paid trial task on real work, and a check of past engagements. Agreements and confidentiality terms are signed before a person joins the bench.

For skills already on the bench there is no search to start: we shortlist from people we have vetted. We give you a date on the first call, and you interview everyone we put forward.

You direct the work, as you would for any member of your team. We handle performance, continuity and cover, and a senior person on our side stays close to the engagement.

Tell us early. We would rather change the person than have you carry a poor fit. If someone has to leave, we replace them from the bench and manage the handover.

It is sized to the work. Some are a few months to get a workstream moving; others run longer as part of a standing team. The duration is agreed up front and can change as the need does.

A rate per person, agreed in writing before anyone starts. You get the figure for your role on the first call. What you are buying is senior capacity without a hiring cycle, and the rate reflects that seniority.

Our people work inside your environment, on your tools, under your policies and access controls. Confidentiality and ownership of the work are written into the agreement before any access is granted, and access is returned when the engagement ends.

We do not publish a number, because the honest answer depends on how many systems and how many teams are in scope, and a figure on a web page would be wrong for most readers. What we do commit to is the shape: a fixed fee agreed in writing before any work starts, no hourly billing, and no change orders for work inside the agreed scope. If the scope genuinely has to change, that is a conversation before it happens rather than an invoice after.

Access and some people’s time, and the specifics depend entirely on your setup. We work that out with you and write it into the proposal before you commit, so nobody discovers a hidden burden halfway through. Two weeks is short enough that we are specific about this rather than optimistic.

For what it promises, yes — and the scope is drawn that way deliberately. Two weeks buys you a scored picture, a ranked gap register, a sized cost view and a sequence. It does not buy you a full financial model or a board paper. We would rather be exact about the boundary than quietly stretch the sprint and miss it.

No, and the deliverables are built on that assumption. Everything is handed over in source and is yours to act on however you choose, including with another firm or entirely in-house. A recommendation you can only act on by rehiring the people who wrote it is not a recommendation, it is a lead magnet.

Sometimes, and we will say so before you buy. Internal assessments usually come unstuck on one of two things: the people doing the scoring are the people being scored, or the cost side was never sized, so there is nothing to take to a budget conversation. If yours has neither problem, we will tell you that on the first call rather than sell you a sprint you do not need.

They score the same six dimensions, and that is deliberate. The free one scores what you tell it, in about five minutes, entirely in your browser. This one scores what we find, over two weeks, against your systems — and adds the two things a self-assessment structurally cannot produce: sized costs and findings someone outside your team will accept.

Your real or production-representative data, approved integrations, representative users, agreed security and access controls, and an expected workload. Customer-facing production deployment is included only when it is approved and inside the agreed scope.

At kickoff, once the access, data, environment and participant prerequisites written into the proposal are available.

Pilots start at $30,000. The exact fee is fixed in writing before work starts and depends on the agreed workflow, systems, integrations, user group and environment.

We are not tied to one model or cloud. Recommendations follow your security requirements, existing technology footprint, workload, quality targets and unit economics.

That is a valid pilot result. You receive the system, evaluation harness, findings and measured economics needed to stop, revise or redirect the use case rather than scale it on optimism.

The engagement is designed for handoff. You receive the source and engagement-specific evaluation assets, subject to the licensing terms of any third-party components used.

The answer depends on the selected enterprise service and contract. Data handling, retention and training terms are documented before build, and the architecture is configured to follow your approved controls.

Whichever fits your data, your existing footprint and your team. We work across Microsoft Fabric, Databricks and Snowflake, and the recommendation is not conditioned on an exclusive vendor relationship — it follows the diagnostic. If the decision is already made, the diagnostic tests it rather than reopening it.

Then that is the finding, and the engagement is re-scoped around what you do need — which is often a smaller amount of fixing than a rebuild. The diagnostic exists so that decision is made on evidence rather than on the assumption that arrived with the budget.

No. It stands up the foundation properly and connects the first sources as the pattern for the rest. A full migration of every source and every workload is the Data Engineering & Platforms practice, and the accelerator is built to lead into it with the pattern already proven.

Your team. The runbook is written for them, they are walked through it, and the first pipelines are built the way every subsequent one should be so the pattern is reusable without us. Where you need more hands, that is a separate conversation about staff augmentation, not a dependency built into the platform.

The fee is fixed in writing before work starts and depends on the number of sources in scope, the state of what already exists and whether the platform decision is made. Work inside that boundary is not billed hourly.

Access to the existing estate and the cloud account the foundation will live in, the people who own the first sources, and whatever architecture or platform decisions already exist. The specifics are written into the proposal before you commit.

It replaces an undefined review with a defined one. Work that is currently waiting for an answer nobody can give gets an answer with a date on it.

Within the agreed scope, yes. Mapping what is in use there is the first stage, and existing systems are brought under the same path rather than grandfathered around it. An enterprise-wide inventory is a practice engagement, not this one.

The NIST AI Risk Management Framework and ISO/IEC 42001 as the structure, your existing policies as the starting point, and the federal and state requirements that apply to you as the constraint. The path is sized for your organization, not lifted from a standard.

We draft it, in your organization's voice, from the review path we design together. Your legal, risk and security owners work through it with us until it is ready for their approval. Your counsel reviews it; this is not legal advice and we do not replace them.

The fee is fixed in writing before work starts and depends on how many AI systems are in scope, how many review functions are involved and how much existing policy there is to build on. Work inside that boundary is not billed hourly.

Access to the people who currently say yes or no, your existing policies, and a rough list of the AI in the agreed scope. The specifics are written into the proposal before you commit.

Neither. The decisions are yours; the AI judgment, the decision method and the challenge in the room are ours. We tell the team which ideas will not survive contact with their data, their controls or their economics, and why — then help them choose among the ones that will.

The people who will fund and be accountable for the first use case: typically the executive team and the function heads most likely to own it. We agree the list in scoping and keep it small enough to decide.

A working session of several hours, with a short pre-read and a few sponsor conversations before it and a readout after. The exact length is set in scoping from the size of the team and how settled the question already is.

Then the readout says exactly what is still contested and by whom, which is the most useful output a divided team can get. Nothing gets funded on the assumption of an agreement that was not reached.

Most workshops are $10,000 to $25,000, fixed in writing before the session. Where in that range depends on the size of the team, the preparation required and whether more than one group is involved. Nothing is billed hourly.

The participant list, a few short conversations with sponsors, and whatever strategy, plans or AI activity already exists. The specifics are written into the proposal before you commit.

Because AI introduces exposure that a programme written before AI arrived does not look for: model behaviour, prompt and data leakage, adversarial input, third-party model dependencies, agents holding credentials. The sprint covers that surface specifically, maps every finding to a control your programme already recognises, and hands the findings to your programme to own. It is built to strengthen the programme you have, not to replace it.

No. It is a posture assessment against recognised frameworks, focused on AI and data. Where a technical test is warranted — a red-team exercise against a specific assistant, say — the readout says so, says what it should target, and that becomes a scoped piece of work in the security practice rather than something hidden inside this one.

Read access to configuration, logs and the data flows in scope, and time with the people who own each system. We do not need to run anything against production, and nothing in scope is changed by us. The specifics are written into the proposal before you commit.

The NIST AI Risk Management Framework and ISO/IEC 27001 as the structure, so every finding lands on a control your auditors already recognise. Where you are held to a sector regime — financial services supervisory guidance, for instance — the register carries that mapping too. We use the frameworks you are measured against, not a proprietary one you would have to translate.

The fee is fixed in writing before work starts and depends on the number of AI systems and data flows in scope and on whether an inventory already exists. Work inside that boundary is not billed hourly.

Whatever inventory of AI systems exists, even an informal one; read access to the configuration and logs of the systems in scope; the people who own them; and the frameworks and regulators you are held to. The specifics are written into the proposal before you commit.