Top 10 AI Consulting Services and Companies for Startups and Enterprises in 2026

The best AI consulting services in 2026 depend on project size. For one workflow and a first project, look at #1–#3, which publish a pilot or proof-of-concept path; for multi-country programs that change operating models, look at #8–#10. This list covers 10 AI consulting services and AI development companies for startups, mid-market teams, and enterprise buyers. Facts were collected from each company's own website, press releases, or filings on September 29, 2026. A company qualified if it sells AI advice together with build or delivery and publishes verifiable proof. Every entry uses the same fields, including a limitation and one question to ask before you sign. WiserBrand publishes this article and is listed first, under the same rules. Ranks 2–10 are grouped roughly from smaller specialists to large enterprise providers and are not a quality score.

Comparison Table: Top AI Consulting Services at a Glance

Which Company Fits Which Scenario

The 10 AI Consulting Services and Companies Worth a Place on Your 2026 Shortlist

Facts below come from each company's own website, press releases, or regulatory filings. Where a company does not publish a fact, the entry says so instead of dropping the field. Certificates are as stated by each company; we did not check certification registries.

#1 WiserBrand - Best for mid-market operations workflows that need a fixed-scope proof of concept

WiserBrand runs AI work as a sequence: a 2–4 week readiness assessment, a six-week fixed-scope proof of concept, implementation, then managed operations. Most of the AI cases published on its site are agents and integrations around systems a company already runs, such as Zoho, Odoo, Zendesk, Magento, and Yardi with QuickBooks, and in those cases a person approves actions that change records. It fits a team with one clear workflow and a budget between $10K and $500K. It does not fit a program that needs thousands of consultants or delivery in many countries at once.

• Type: Implementation-led AI consultancy: advisory, build, and managed run.
• Delivery scope: Advisory (AI Strategy Consulting, AI Readiness Assessment), proof of concept, build (AI agents, AI integration, custom AI), and run (Managed AI).
• Geography: Headquarters at 1120 Avenue of the Americas, New York; R&D center in Buenos Aires, Argentina. The About page lists 180+ professionals.
• Engagement and pricing: Rapid AI Proof of Concept: six weeks, one use case, one primary data source, $30–75K. Single-workflow agent pilot: $10–40K, with the price fixed before build starts. Implementation: $120–500K. Managed operations: $10–40K per month.
• Compliance and certifications: No ISO 27001 or SOC 2 certificate published. Service pages state that controls are mapped to GDPR/CCPA, HIPAA, and SOX/FINRA requirements.
• Post-launch ownership: Managed AI covers monitoring, drift detection, and 24/7 incident management, with SLAs reviewed quarterly; numeric SLA values are not published. Ownership of custom code, configurations, and model artifacts is "defined in the project agreement."
• Proof (results as reported on the case pages): Supplier quote verification for a US construction company: built in 6 weeks, 68% faster quote review, 92% first-pass extraction accuracy on about 118 quotes a month. AI HR assistant: 60% faster CV review, about 20 hours saved per week. Custom MCP server for Magento: 80% less time on parts search. AI legal suite for a bank: 88% less processing time in a suite that handles 1M+ contracts a year. Inc. 5000, Google Partner, and Clutch designations are listed on the homepage without years.
• Limitation: Most clients are anonymous, and dollar figures on several cases are labeled "estimated annual value" without the method shown. There is no ISO or SOC 2 certificate. It is not a strategy-only shop: the advisory work is built around a follow-on pilot.
• What to check: Which of your delivered cases used the same systems as ours, and what does the proof-of-concept contract say about who owns code, prompts, and evaluation sets?

#2 InData Labs - Best for startups and mid-market teams that want a PoC or MVP built and then maintained

InData Labs publishes price bands, which makes its quotes easy to compare when the budget is small. It sells AI consulting as the first step toward a build: the AI consulting page describes a readiness assessment, an MVP, a proof of value, and ongoing maintenance, and a December 2025 blog post addresses startups directly.

• Type: Custom AI and data science development firm with an advisory front end.
• Delivery scope: Advisory (readiness assessment with report, roadmap, and ROI projection delivered in 1–2 weeks), build (data science, LLM and RAG, agents), deploy, and run (support tiers for monitoring, retraining, and performance reviews).
• Geography: Headquarters in Nicosia, Cyprus; the contacts page lists Miami as a sales office and a third office in Vilnius. The team works remotely across 24 countries. Team size is stated as 70+ on one page and 80+ on others.
• Engagement and pricing: The homepage lists $15–50K for a PoC or MVP, $75–200K for mid-complexity solutions, and $400K–$2M+ for enterprise platforms. The AI consulting page states $10K to $250K+, so confirm which band applies to your scope.
• Compliance and certifications: No ISO 27001, SOC 2, or HIPAA certificate is published. The privacy policy references GDPR and a named data protection officer, and the AI consulting page states alignment with the EU AI Act.
• Post-launch ownership: Support tiers are described. The site names no code or IP ownership terms and no SLA values.
• Proof: Clients named on the About page include Wargaming, GSMA, Flo, and Captiv8. The homepage claims 150+ delivered projects and a 4.9/5 Clutch rating, plus a 30% operational-efficiency gain from LLM analytics and 2x sales conversions from an AI virtual assistant (company-reported).
• Limitation: The published price bands contradict each other across pages, and a team of 70–80+ is small for the $400K–$2M+ tier it advertises.
• What to check: Send the master services agreement showing who owns code and model weights at delivery, and the SLA for the maintenance tier.

#3 Markovate - Best for document- and drawing-heavy operations that want a pilot on their own data

Markovate's pitch is a pilot on your own data: its homepage offers a "focused pilot in 4–6 weeks." It builds agentic and generative AI for construction, manufacturing, healthcare, insurance, and real estate, and its AI consulting page lists feasibility studies, proof of concept, and audits of existing AI.

• Type: Product builder with vertical solutions.
• Delivery scope: Advisory, build, and deploy. Ongoing support is offered, but its terms are not on the site.
• Geography: Offices in San Francisco, Schaumburg (Illinois), Toronto, and Gurugram (India); no headquarters is designated. The About page lists a core team of 50+ (undated).
• Engagement and pricing: Pilot first, then full delivery. Pricing and minimum engagement do not appear on the site.
• Compliance and certifications: The Trust Center states ISO/IEC 27001:2022 and ISO 9001:2015. No SOC 2 is listed. HIPAA and GDPR are worded differently across pages ("ready," "best practices," "compliant"), so treat them as claims, not certifications.
• Post-launch ownership: The Trust Center states 99.5% availability across enterprise deployments with 24/7 monitoring, and says data and custom models built under formal agreements remain with the client. Code and IP terms are not stated.
• Proof: NVMS: 70% less photo inspection time. Aisle 24, a Canadian retailer: 3x sales. Anonymized results include insurance claims processed 40% faster (company-reported). Partner statuses listed: AWS, Microsoft Solutions Partner, Google Cloud, with no tiers stated.
• Limitation: No pricing or headquarters, and most case results are anonymized. The homepage says 50+ AI projects delivered and the About page says 200+, so the counts do not agree.
• What to check: The ISO 27001 certificate scope and certifying body, which team (India or North America) would build the project, and who owns code, models, and IP after the pilot.

#4 LeewayHertz - Best for custom generative AI and agent builds

Check the contracting entity first. LeewayHertz builds RAG systems, chatbots, and workflow agents and offers AI strategy consulting ahead of the build, but it has been a subsidiary of The Hackett Group since September 2024: Hackett announced the acquisition on September 16, and its SEC filing records the purchase closing on September 23, 2024 for $7.8M.

• Type: AI and software development firm with a consulting practice, a subsidiary of The Hackett Group (NASDAQ: HCKT).
• Delivery scope: Advisory, proof of concept, build, and deployment with MLOps consulting. Run appears only as prompt maintenance; no managed offering is described.
• Geography: The only office address on leewayhertz.com is in Gurugram, India. The parent, Hackett, is headquartered in Miami. LeewayHertz's own headcount is not stated; Hackett reported 1,503 associates at the end of 2025 for the whole group.
• Engagement and pricing: Dedicated development team, team extension, or project-based. No prices on the site.
• Compliance and certifications: The About page states ISO/IEC 27001:2022, ISO/IEC 42001:2023, SOC 2 Type II, HIPAA, and GDPR. Certificate scope, certifying body, and the legal entity holding them are not stated.
• Post-launch ownership: No IP or SLA terms appear on the site. The ZBrain agent platform is owned by Hackett.
• Proof: The site claims 100+ AI projects and 30+ Fortune 500 clients. Named work includes an LLM service-desk chatbot for Rackspace built on its ServiceNow knowledge base, with qualitative results only, and testimonials from Siemens and O'Reilly Auto Parts.
• Limitation: Independence and roadmap now depend on Hackett; headcount, IP, and SLA terms are not stated, and case outcomes are mostly qualitative.
• What to check: Which legal entity signs the contract, who owns the IP of agents built together, and whether those agents can run without a ZBrain license.

#5 Neurons Lab - Best for mid-market financial services firms deploying agentic AI

A narrow focus is Neurons Lab's selling point: agentic AI for financial services, plus an AI training arm for leadership and technical staff. Projects run in four phases (discovery, pilot, production, expansion), and the FAQ says proofs of concept typically take 2–4 weeks.

• Type: Vertical specialist (financial services) with a training and enablement arm.
• Delivery scope: Advisory, build, and deploy, plus an embedded "Continuous AI Delivery" offer. Run terms are not stated.
• Geography: London (named as headquarters on the site) and Singapore. A June 2026 release lists presence in the US, UK, Europe, and Singapore. The About page states 50+ AI engineers, architects, and analysts across Europe, and the careers page describes a separate talent network of project-based contributors. Founded 2019.
• Engagement and pricing: Project-based. The FAQ gives $20–50K for short focused engagements such as rapid PoCs, and $200–750K for multi-month programs.
• Compliance and certifications: No ISO 27001 or SOC 2 certification for the company is published. The FAQ says solutions align with GDPR, HIPAA, and PCI DSS, and offers on-premises or private-cloud deployment.
• Post-launch ownership: Handover deliverables are listed: technical documentation, architecture diagrams, infrastructure as code, training manuals, and operational dashboards. Code ownership, IP, and SLAs are not stated.
• Proof: HSBC (executive enablement for 50+ leaders), Visa (marketing LLM tooling across 9+ markets), and AXA (named in the June 2026 AWS announcement). AWS AI Competency in the Agentic AI category announced June 2, 2026. The company claims 100+ implementations since 2019.
• Limitation: Financial services focus, and startups are not addressed. The cheapest published engagement is about $20K.
• What to check: Who owns agent code, prompts, and fine-tuned assets at handover, and which of the 50+ people would work on the project as core employees versus talent-network contractors.

#6 Azati - Best for regulated or legacy-heavy enterprises that need AI built into existing systems

Azati is engineering first and AI second: AI consulting is one line inside a broader engineering offer, next to an "AI MVP Development & Rapid Prototyping" service that promises first users 6–8 weeks after scoping.

• Type: Custom software engineering firm with an enterprise AI practice.
• Delivery scope: Advisory (AI strategy and a data readiness assessment listed at 2–4 weeks), build, and run (Managed AI with model monitoring and prompt-drift detection under defined SLAs, plus retainer support).
• Geography: US office in Livingston, New Jersey; Warsaw, Poland is the European base and main development center. The company profile states 300+ in-house specialists and a 2002 founding.
• Engagement and pricing: Dedicated development center (minimum 8 engineers), project-based (2–30+ engineers), or staff augmentation; fixed price or time-and-materials. No AI consulting price ranges. One blog post cites about $10,000 for a recommendation-system MVP.
• Compliance and certifications: The site says ISO 27001, 27701, and 9001 accreditation is in progress with a 2026 target. GDPR protocols and "HIPAA-aware" engineering are stated. No SOC 2 is listed.
• Post-launch ownership: The company profile states that on project completion "all resulting deliverables become your exclusive property." Managed AI SLA values are not stated.
• Proof: Client logos include Shell, Clarivate, and MedPro Group, but the named clients are not tied to specific case studies. Anonymized results: a patent-search platform processing 50M+ documents with 72% less manual work, and oil-and-gas document processing with 70% less manual correction.
• Limitation: ISO certifications are pending, results are anonymous, and the dedicated-center model needs at least 8 engineers.
• What to check: Which reference clients can speak to an AI project like yours, where the ISO 27001 audit stands today, and what the Managed AI response times are.

#7 Fractal Analytics - Best for large enterprises in CPG and retail, TMT, healthcare, and banking

Fractal sells enterprise AI as services plus its own products, including the Cogentiq agentic AI platform. A red herring prospectus preceded its 2026 listing on the NSE and BSE and is the source of several figures below.

• Type: Enterprise AI and analytics services company with proprietary products.
• Delivery scope: Advisory, build, and deploy. The site describes no managed-service offering; the prospectus lists subscription and licensing among its pricing models.
• Geography: Registered office in Mumbai; the contact page names New York as global headquarters. Other offices span the US, India, Australia, Canada, London, Eindhoven, Singapore, Dubai, and Kyiv. It reported 6,029 employees on June 30, 2026, and 67.4% of segment revenue came from the Americas in Q1 FY27.
• Engagement and pricing: The prospectus lists four pricing models: fixed price, subscription and licensing, output-based, and time-and-materials. Rates and minimum engagement are not stated.
• Compliance and certifications: The prospectus says the security framework is "aligned with" ISO 27001, PCI DSS, and SOC 2 Type II. That is weaker than a certification, and no certificate list appears on the site.
• Post-launch ownership: The prospectus says Fractal protects its own IP through client agreements. IP terms for client deliverables and SLA values are not stated.
• Proof: Clients named in the prospectus include Citibank, Costco, Franklin Templeton, Mars, Mondelez, Nationwide, Nestle, and Philips. Fractal states it worked with 10 of the 20 largest CPG companies as of March 31, 2025. AWS Premier Tier Services Partner (February 2025) and Preferred Services Partner in the Claude Partner Network (July 7, 2026).
• Limitation: Enterprise scale only. The top 10 clients made up 51.8% of segment revenue in Q1 FY27, and startups and mid-market are not a stated focus. Most case studies are anonymized.
• What to check: Who owns the models, agents, and prompts built on Cogentiq or Claude for you, and whether a SOC 2 or ISO 27001 certificate with scope exists for the delivery team that would touch your data.

#8 QuantumBlack, AI by McKinsey - Best for large enterprises tying AI to operating-model and workforce change

QuantumBlack is McKinsey's AI and engineering arm, part of the firm since 2015. A published example of its scope is the Merck engagement: a joint team of 80+ people cut the first draft of clinical study reports from 180 hours to 80 hours and halved draft errors (company-reported).

• Type: Strategy firm's AI and engineering arm.
• Delivery scope: Advisory, build, deploy, and run. Its Labs page describes 20+ AI products and 140+ use-case accelerators that can be client-managed, managed by QuantumBlack, or delivered as a service.
• Geography: No headquarters is stated. The people page lists 25 office locations, and McKinsey overall has offices in 130+ cities. Current headcount is not stated.
• Engagement and pricing: Multidisciplinary teams work alongside client teams in agile sprints. Pricing and minimum engagement are not stated.
• Compliance and certifications: None stated for QuantumBlack as a service.
• Post-launch ownership: A live operations team monitors and improves deployed solutions, then either transitions them to the client or continues. IP terms for client work are not stated. Kedro and Vizro are open source.
• Proof: Merck (above), Deutsche Telekom (a capability engine to upskill 8,000 field and call-center agents), and KPN (an agentic engine for customer care). ING, Aviva, and Freeport-McMoRan appear on the case-study list.
• Limitation: Pricing is opaque and the case base is very large enterprises. The transformation scope suggests long, senior-heavy engagements; that is our inference, not a published fact, and a small team with one bounded build may find it a poor fit.
• What to check: Which QuantumBlack products and accelerators you would depend on after handover, whether they are licensed or client-owned, and who runs them after the transition period.

#9 Infosys - Best for large enterprises embedding AI in big application and IT-operations programs

Infosys fits when AI is one part of a multi-year application and operations program; for a standalone build, its size becomes overhead. It packages AI work as Infosys Topaz and Topaz Fabric: the Topaz page cites 12,000+ AI assets, 10+ AI platforms, and 150+ pre-trained models, and a May 2025 press release describes over 200 enterprise AI agents built with Google Cloud.

• Type: Global IT services and consulting firm with an AI-first services and platform layer.
• Delivery scope: Advisory (Infosys Consulting), build, deploy, and run. Recent managed-services deals include Knorr-Bremse and GlobalFoundries.
• Geography: Headquarters in Bengaluru. Its 20-F filing lists 290 locations in 59 countries and 328,594 employees at March 31, 2026, of which 258,502 were in India.
• Engagement and pricing: Fixed-price, fixed-timeframe projects were 54% of FY2026 revenue. The 20-F says output-based and transaction-based pricing is offered in some situations. Minimum engagement is not stated.
• Compliance and certifications: The certifications page lists ISO 27001:2022 (valid to December 17, 2027), ISO 42001:2023 (valid to March 29, 2027), ISO 27701, CMMI Level 5, and SOC 1 Type II. SOC 2 is not named.
• Post-launch ownership: The Knorr-Bremse and GlobalFoundries deals are described as AI-led managed services. IP terms for client deliverables are not stated.
• Proof: A testimonial on the Topaz page cites fraud worth about Rs 42,000 crore (about $5 billion) uncovered at India's GST Network. Named 2026 deals include Knorr-Bremse, Crocs, Sentara, and GlobalFoundries, without published outcome numbers.
• Limitation: With 328,594 employees, most public evidence is announcements rather than measured outcomes at named clients, and you will need to confirm who leads your team.
• What to check: Which of the 12,000+ assets apply to your stack, what measured production results exist at a comparable client, and what happens to IP and pricing under an outcome-based contract if your AI volume rises or falls.

#10 Cognizant - Best for large enterprises in healthcare, financial services, and life sciences

In its 10-K Cognizant calls itself an "AI builder," and it sells AI through platform families under the Neuro AI name plus an Agent Foundry service. It had 356,700 employees at June 30, 2026 and $21.1B in 2025 revenue.

• Type: Global IT services and consulting firm.
• Delivery scope: Advisory, build, deploy, and run. The 10-K lists consulting, application development, systems integration, application maintenance, infrastructure, and business process services. It completed the acquisition of Astreya, an IT managed-services provider, on June 22, 2026.
• Geography: Headquarters in Teaneck, New Jersey. About 90% of delivery-center square footage is in India, per the 10-K.
• Engagement and pricing: Fixed-price contracts were about $10.0B of $21.1B in 2025 revenue (about 47%); the rest is time-and-materials, transaction-based, or volume-based. Minimum engagement is not stated. A Microsoft Copilot accelerator runs eight weeks for up to 500 users; it is a product rollout, not a benchmark for AI consulting engagements.
• Compliance and certifications: The 10-K states ISO 27001, UK Cyber Essentials Plus, and ENS. A company-wide SOC 2 or ISO 42001 is not stated.
• Post-launch ownership: IP terms for client deliverables and Neuro accelerators are not stated.
• Proof: Travelport, with Anthropic, announced May 27, 2026, with first customer-facing capabilities expected in 2026. Agent Foundry claims content approvals cut from four weeks to four minutes, without naming the client. The Neuro platform pages publish few metrics or named clients.
• Limitation: Most quantified AI outcomes on its pages are unnamed-client claims or Cognizant's internal use.
• What to check: Name a referenceable client in your industry running a Neuro or Agent Foundry solution in production with a measured result, and who owns the agent code, prompts, and context layer if you leave.

How WiserBrand's AI Consulting Services Map to Each Stage

An AI consulting engagement usually moves through five stages. This is how WiserBrand's published services and cases line up with them. Dollar values on cases are WiserBrand's own estimates, as labeled on each case page.

AI Consulting for Startups: What Changes

AI consulting for startups differs from the enterprise version. The budget is smaller, the goal is usually one working product or workflow rather than a transformation program, and the risk is paying for strategy work that never ships. Among the ten companies above, the ones that publish a startup-sized entry point are:

• WiserBrand: single-workflow agent pilot at $10–40K, fixed price before build starts.
• InData Labs: PoC or MVP from $15–50K, with a blog post written for startups.
• Neurons Lab: focused engagements at $20–50K, though its focus is financial services.
• Azati: AI MVP with first users 6–8 weeks after scoping; price not stated.
• Markovate: pilot in 4–6 weeks; price not stated.

QuantumBlack, Infosys, Cognizant, and Fractal do not publish minimum engagements and describe enterprise clients. Startups can still buy from them, but expect to request a price range up front. Among the AI consulting startups and small firms on this list, Neurons Lab (founded 2019, 50+ engineers), Markovate (core team of 50+), and InData Labs (70–80+) are the smallest of the ten by stated headcount. Small firms often start faster and put delivery in fewer people, so ask who those people are.

For a first project, published entry points among these vendors run $10K–$75K, with a six-week proof of concept at the top end. We suggest comparing any quote far above that with a second vendor before a working prototype exists. That is our judgment, not a benchmark.

How to Choose the Best AI Consulting Services Without Backing the Wrong One

Analyst estimates of the market differ by publisher. Fortune Business Insights projects the AI consulting services market at $11.91B in 2026, growing to $73.89B by 2034 (25.6% CAGR), while Zion Market Research puts 2024 at $8.75B. Treat any single figure as one firm's estimate.

Delivery risk is the better reason to vet carefully. A Gartner survey of 644 respondents, fielded in Q4 2023, found that on average only 48% of AI projects make it into production and that the move from prototype to production takes 8 months (Gartner, May 2024). Gartner also predicts that over 40% of agentic AI projects will be canceled by the end of 2027 because of escalating costs, unclear business value, or inadequate risk controls (Gartner, June 2025). One is a survey and the other a prediction, not a measured failure rate, but they point the same way.

If the vendor will serve EU customers or process EU data, check regulation timing. The rules for general-purpose AI models applied from August 2, 2025. The AI Omnibus, which entered into force on July 27, 2026, moved the high-risk obligations for areas such as employment and education to December 2, 2027, and for AI built into regulated products to August 2, 2028 (European Commission).

Vendors use the terms below differently, so ask which one a price covers. A proof of concept tests whether one use case works on your data, without live users. A pilot runs the workflow live with a limited audience. An MVP is a product you can release.

Not every team needs a consultant. If your engineers already own a well-defined use case, building in-house or using a software vendor's built-in AI may be enough; consulting helps most when scoping, data readiness, or delivery capacity is the gap.

Use this table to narrow the ten and to structure the first call with each candidate:

Next Step: Pick a Shortlist and Test It

Use the scenario table to pick two or three companies and send each the same eight questions. Ask for three things in writing: a price range for the first six weeks, ownership terms, and a success metric with a go/no-go date. If your workflow is a single, bounded one and you want a comparison quote, WiserBrand offers a 2–4 week readiness assessment and a six-week proof of concept. Request a 15–20 minute exploratory call.

FAQ

How much do AI consulting services cost?

Published entry points start at $10K. WiserBrand lists $10–40K for a single-workflow agent pilot and $30–75K for a six-week proof of concept, InData Labs lists $15–50K for a PoC or MVP, and Neurons Lab lists $20–50K for focused engagements. Larger work costs more: WiserBrand publishes $120–500K for implementation, and Neurons Lab publishes $200–750K for multi-month programs. These are prices the companies publish themselves, and we found no independent benchmark. QuantumBlack, Infosys, and Cognizant do not publish prices, so ask for a fixed price before any build starts.

How long does an AI consulting engagement take?

A readiness assessment takes about 2–4 weeks (WiserBrand). A proof of concept takes 2–6 weeks: WiserBrand runs a six-week fixed scope and Neurons Lab quotes 2–4 weeks. Markovate quotes a 4–6 week pilot, and Azati lists first users 6–8 weeks after scoping for an MVP. For the move to production, Gartner's 8-month average (above) is the planning reference.

What should a proof of concept prove before we pay for production?

It should show a measured result against a baseline agreed in advance, on your data, with a go/no-go date. WiserBrand's proof-of-concept page lists business KPIs measured with confidence ranges, a risk register, and a 90-day pilot plan with a production budget as deliverables. Whatever the vendor, write the success metric, the measurement window, and the go/no-go date into the contract before build starts.

Who owns the code after an AI consulting project?

Ownership terms are rarely published. Azati states that deliverables become the client's exclusive property, WiserBrand says ownership is defined in the project agreement, and Markovate says data and custom models stay with the client. The other seven companies publish no ownership terms. Do not sign until the contract assigns code, prompts, evaluation sets, and model artifacts to you.

Should a startup choose a boutique firm or a large consulting company?

Use scope as the decision rule. For one workflow with a published pilot or proof-of-concept path, choose a specialist. For a multi-country program that changes operating models and workforce roles, choose a large provider. The rule is our judgment, not a published benchmark, so test it against two or three quotes.

Which certifications should an AI consulting company hold?

If the vendor will touch regulated or customer data, require a current ISO 27001 or SOC 2 Type II certificate with its scope and expiry date. Among the ten, Markovate, LeewayHertz, Infosys, and Cognizant state ISO 27001, and Infosys also lists ISO 42001, the AI management system standard. Azati's ISO certifications are still in progress.