The Total Economic Impact™ Of Microsoft Azure AI

Business Growth

• Evidence and data.

  Azure AI spurred business growth for the interviewee’s organizations in a variety of ways depending on their use case and products utilized. Customers enjoyed increased revenue through enhanced business volume, improved customer acquisition and retention, and the development of new products and services. Specific examples of customers’ journeys from Azure AI to revenue growth included:

• Automation improving process scalability and increasing work volume.

  Process automation powered by Azure AI significantly improved operational efficiency and increased the organizations’ capacity to scale up the volume of their work, or output, in the section of the business that applied AI. This automation improved the efficiency and quality of service delivery to clients or customers and helped organizations support more customers or produce more of a product, facilitating additional sales and increasing revenue.

  Azure OCR helped the aviation organization grow their document processing, indexing, and classification services, allowing them to serve additional clients. The senior principal of digital at an aviation firm said, “[Azure OCR] allows us to immediately increase our scale without incurring additional resources on our end to scale.” They noted that they processed six to eight times as many documents with Azure AI as they could with their old open source OCR software. Before their investment with Azure AI, they could only support one or two major clients per year, but that capacity increased to six within a few years.

  The energy organization used Azure AI for remote asset operations and monitoring at a scale that they were unable to reach without AI technology. The principal AI architect explained: “There isn’t enough human SME capacity to do [asset monitoring] in any other way. They’ve tried doing it by eyeballing real-time analytics, and they physically can’t scale to look at all of that. … We’re talking 6 million custom models to cover all the equipment in the company. With Cognitive Services, you can roll that out very quickly across one base implementation where you don’t have to custom craft all these models.”

  A partner at the professional services organization described using Azure AI to audit financial transactions, interpret data, and run document intelligence for clients. They noted that their tax solution with Azure AI alone drove millions of Euros of business for their organization, and using Azure AI was essential for them to provide the services at the scale they needed.

• Improved customer acquisition and retention.

  Using Azure AI for customer-facing use cases improved the quality of their services, resulting in more clients and customers being acquired and retained.

  The partner at a professional services organization said Azure ML reduced their rate of errors in their transaction monitoring. This helped their clients avoid large fines or false accusations, and improved their customer satisfaction.

  The technology organization worked largely with manufacturing companies, most of whom worked with Microsoft as a partner or vendor. Offering a solution built on Azure helped the technology organization gain market credibility and made their product or service offerings more appealing for customers, their existing technology stack would integrate well with the offerings.

  The aviation organization improved their customer acquisition with a faster onboarding process and customer retention by reducing errors, and enabling faster delivery of their document processing services. Azure AI enabled more effective set up of their customer document templates and data pipelines, reducing their customer onboarding timeline from six to 12 months to one to three months. Their prior process offered less visibility into their pipeline and less consistency in their work, and was unable to guarantee customers would receive their processed documents within reasonable timelines. Azure AI enabled them to offer time estimates of 20 minutes to one hour, from their previous estimates of three to four hours. The senior principal of digital described: “Before, when the customer asks us when it arrives, it can be one hour, or four hours. With the new process, customers are really happy because they can immediately see the impact.”

• The development and launch of new products and services.

  Azure AI enabled interviewee's organizations to develop new products and services, resulting in new revenue streams.

  The technology company developed two new products with Azure AI: One was an industrial analytics and AI suite, where they help customers develop their own AI applications with Azure, and the other was an asset performance management (APM) solution that they sell directly to customers. The APM solution utilizes AI applications for anomaly detection, predictive maintenance, and risk analysis. The organization’s technology head of digital portfolio management said: “We help our customers build their applications using the AI/ML capabilities and contextualized data and our domain knowledge. We bring all that together to provide solutions to our customers.”

  The professional services organization then developed and licensed a documenting intelligence and voice intelligence platform solution built on Azure AI, which they sell as a managed services technology.

• Modeling and assumptions.

  This section explains how the modeling is done:

• The composite utilizes AI to increase work volume for a team with a $150 million revenue stream that serves business customers.

• For Your organization, the size of the revenue stream before the investment in Azure AI is x.

• With Azure AI, the team’s total volume of work output increases by 75% in Year 1, 125% in Year 2, and 150% in Year 3, allowing them to serve more customers and generate added revenue.
• Thirty-three percent of the increased work output can be directly attributed to Azure AI. Attributing one-third of the outcome to people, process, and technology, Azure AI makes up the entire technology piece.
• The organization develops a new product based on Azure AI for business customers. The product is launched midway through Year 2 — it generates $24 million in revenue in Year 2, and $48 million in Year 3.

• Your organization plans to develop new products or services based on Azure AI after x of usage. This products and services might generate x in revenue in Year 1, x in revenue in Year 2, and x in revenue in Year 3.

• Fifteen percent of the new product revenue can be directly attributed to Azure AI. This is a lower attribution percentage because the technology itself played a smaller role in the total design and sales success of the new product.
• The composite has an operating profit margin of 10%.

• Your organization has an operating profit margin of x.

• Risks.

  The expected financial impact is subject to risks and variation based on factors including the organization’s:

• Azure AI scope of deployment and use cases.
• Type of revenue source, its annual number of sales and revenue per sale, and operating profit margin.
• Level of internal investment in model development, training, and fine-tuning.

• Results.

  To account for these risks, Forrester adjusted this benefit downward by 15%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $12.7 million.

  For Your organization, the three-year, risk-adjusted total PV might be x To account for risks and based on your planned Azure AI usage, this benefit is adjusted downward by x .

Spending Optimization

• Evidence and data.

  AI and ML analyses and predictive models can identify patterns or insights that humans cannot — enabling businesses to make more informed decisions and identify new growth or optimization opportunities. Azure AI customers reported a variety of cost optimization use cases, from minimized energy consumption to optimized inventory expenditure and storage. Azure ML increase model deployment, accuracy, and quality for organizations that were already using internal ML modeling, allowing them to further minimize their expenses. Improved monitoring and standardized production also ensured models stayed accurate or increased accuracy over time. Customers described realizing cost savings with Azure AI:

• Reduced energy consumption.

  The principal AI architect at the energy organization described using a model to predict and optimize the best shipping routes, timing, and speed of large shipping vessels to use the smallest amount of bunker fuel, thus generating large cost savings. They explained: “Those vessels generally cost about a quarter of a million dollars a day, that can go up to $1.5 million a day depending on how tight the market is. A lot of that cost can be optimized by optimizing whether it’s a port cost or a bunker cost. They’re pretty big numbers.”

• Reduced maintenance costs.

  The technology and energy organizations used Azure AI for remote asset monitoring and predictive maintenance to minimize costly breakages and downtime of large equipment. This was possible by being able to anticipate and fix problems before equipment failed. Preventing equipment failures greatly reduced the equipment’s lifetime cost of maintenance labor and replacement parts. The technology organization reported a 10% reduction in maintenance costs for its APM customers.

• Optimized materials or commodities purchases.

  Using Azure AI, the retail organization was able to more accurately predict the exact amount of raw material to purchase, and the optimal time to purchase materials to receive the lowest possible price. The program manager of the retail organization said, “The initial target set by the commodities team was to beat the spot price by 1%, which doesn’t sound like a lot. But when you translate it into the volume spent, it does become quite a lot. … We beat the 1% after about five months.” They evaluated that they could have saved between 8% and 10% of their overall spend by reducing waste from having too much raw material in inventory over the last 10 years if they had been applying the Azure AI models. They also enjoyed greater returns on their budget by having more accurate predictions on the amount of money they needed in the future.

• Modeling and assumptions.

  Forrester modeled the impact for the composite organization assuming:

• The composite develops AI applications to optimize costs equaling 1% of its total operating costs in Year 1, which grows to 2% of its operating costs in Year 2.

• For Your Organization, the operating costs that might be affected by the investment Azure AI are x in Year 1 x in Year 2, and x.

• The organization reduces its overall expenditure in the affected cost streams by 4% in Year 1, 5% in Year 2, and 7% in Year 3.
• Fifty percent of reduced spending can be directly attributed to Azure AI. The predictive and optimization insights from Azure AI provides the biggest enabler of change, although people and process factors are still required to apply the insights effectively to realize cost savings.

• Risks.

  The expected financial impact is subject to risks and variation based on factors including the organization’s:

• Scope of deployment and use cases with Azure AI.
• Types of costs faced and opportunities for optimization and cost cutting.
• Level of internal investment in model development and fine-tuning.

• Results.

  To account for these risks, Forrester adjusted this benefit downward by 20%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $16.1 million.

  For Your Organization, the three-year, risk-adjusted total PV might be x. (To account for risks and based on your planned Azure AI usage, this benefit is adjusted downward by x.)

Manual Process Automation

• Evidence and data.

  Process automation not only helped grow revenue, but also allowed employees to focus on higher-value work, thus increasing their productivity. The type of process automated and the level of automation varied across organizations, ranging from account or transaction monitoring and fraud detection, demand forecasting, remote operations, equipment monitoring, document intelligence, and data verification. Examples of customers saving their employee’s time with Azure AI included:

• Shortening the time required to complete manual work.

  The partner working in professional services described highly scalable savings from automating most of their auditing process with Azure AI and reducing the number of documents auditors had to manually review. They said, “Doing Document Intelligence on [our auditor and tech] platforms and using ML to run audit procedures is driving massive savings.” An implementation team of seven people at the aviation organization that builds and refines document templates for new customers found that Azure’s sophistication and accuracy saved them time previously spent tuning and retuning models to fit templates. The senior principal of digital shared: “Typically, it would take around three to four days for one template. Now it’s within hours because the results are pretty remarkable. …[With] the time they needed to focus on doing the template tuning, [they can] now focus on other things that is more value-added to our customers.”

• Eliminating the need for human involvement in processes.

  The energy organization uses Azure Cognitive Services’ knowledge mining and vision capabilities to take highly unstructured data and combine it with structured operational data, to automatically build and update digital twins of large equipment. The principal AI architect in this organization described an immense improvement in quality with Azure AI: “A 3D construction model to build a massive asset might have 70 to 80 million component parts. Generally, [over] 20% will be incorrect, so you don’t have confidence to translate it directly into the digital twin, and that requires a huge army of people to inspect or check it against our reference data and work out all the corrections that need to be done. That’s what we are proving can be delegated to Azure AI.”

  The digital twins enable remote operations, which saved the energy organization $117 million a year by eliminating the costs of training people and sending them to remote areas to verify and manage the equipment. The principal AI architect said, “We’re so confident on high fidelity digital twins. Everyone in the supply chain can do it remotely.”

  The retail organization’s procurement commodities team used Azure ML to take over forecasting of price and demand for raw and packaging materials. This work was previously done by seven or eight very senior-level financial research analysts who were experts on the different materials. The research analysts either left the organization or their roles evolved to undertake new tasks.

• Modeling and assumptions.

  This section explains how the modeling is done:

• The composite applies AI to automate manual processes done by a team of 16.

• Your Organization might apply AI to automate manual processes done by a team of .

• Azure AI reduces the manual effort required by 80%.
• Fifty percent of the automation can be directly attributed to Azure AI, reducing the effective required project FTE employees by six.

• At Your Organization, the effective required project FTE employees might be reduced by x.

• The project team members had fully burdened annual salaries of $119,700 on average.

• Risks.

  The expected financial impact is subject to risks and variation based on factors including the organization’s:

• Azure AI scope of deployment and use cases.
• Types of costs faced and opportunities for optimization and cost cutting.
• Level of internal investment in model development and fine-tuning.
• Number of project team members and their compensation amounts and structures.
• Unique business requirements and complexities that may reduce potential time savings.

• Results.

  To account for these risks, Forrester adjusted this benefit downward by 10%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $1.6 million.

  For Your Organization, the three-year, risk-adjusted total PV might be x. (To account for risks and based on your planned Azure AI usage, this benefit is adjusted downward by x.

Operational Efficiency

• Evidence and data.

  Azure offers organizations an efficient and scalable way to develop, train, and validate ML models, thus improving the productivity of data scientists working on ML activities. Customers stated that Azure ML simplified data infrastructure management, modernized template and ML model development, training, validation, deployment, and fine-tuning. Engineers gained access to AI and ML methods and frameworks, accelerating production and overall lifecycles. Ultimately, Azure AI enabled the deployment of more models and applications.

• Multiple interviewees mentioned that consolidating and integrating their data and applications with one provider simplified processes. The partner at the professional services firm described, “Having [all our data] in one player with the hyperscalability makes things much easier.”
• The aviation organization was previously working with another OCR tool that required heavy customization, incurring a lot of time to set up and train the solution. The senior principal of digital in this organization noted: “Before [Azure OCR], it took us three months with six engineers to [train an instance] with the open source process. With [Azure OCR], we are cutting down from three months of engineering time with six people. [Now,] we have one engineer who takes two weeks.” They also noted improved productivity in maintenance and implementation.
• The principal AI architect at the energy organization described efficiencies of Azure AI’s use with other products from the Microsoft ecosystem. They said: “With Cognitive Services and the workflow with things like Power Apps and MS Flow, you can quickly and easily operationalize something as a business workflow to replicate what the business was doing. Over time, you can industrialize it by moving more of it service side so you can scale it.” They noted that Azure Cognitive Service’s multivariate anomaly detection model reduced the full lifecycle of development through production to deploy each new piece of modelled equipment from six weeks to one week. They believed that the next stage of automation would reduce that timeline to a full lifecycle in one day.
• The program manager in retail worked to create processes with Azure AI that were repeatable and efficient. They told Forrester: “We’ve come up with a very robust, tried, and proven methodology where we can onboard pretty much any new use case with a consistent approach and come out with an initial model in a very short period of time.” They also noted that they tried to set up as much of a hands-off solution as they could, automating alerts to remove human oversight. They described: “We have automated outlier detection on the raw data ingestion, on either missing or inaccurate data. When we do the modeling, we have outlier detection in terms of the model accuracy. If we see a significant drop in the model accuracy based on new data, we have alerts, so one of our end-users on the procurement team should already have been alerted and reacted to it in order to ensure that operations are not introducing incorrect numbers.”

• Reduced error rates.

  Azure AI outperformed previous predictive, identification, or processing outputs, whether they were performed by humans or other technologies.

  The partner at the professional services organization described large time savings by reducing the rate of false fraud identification in their transaction monitoring services. They said: “Where we utilize ML, we have seen reductions in the false positive [rate] of 80%. You can imagine the time that you can save because if you perform a false positive handling — it consumes a lot of resources.”

  The senior principal in aviation explained that using Azure’s OCR in replace of their prior OCR tool reduced their document error rate from 4% to 1.5%. Customers generate support tickets for errors or if documents take longer than three or four hours to arrive. The organization saw a direct impact of the reduced error rate when Azure AI decreased that ticket volume by 40%.

  The retail organization’s program manager described that while some of their spending was previously misclassified due to data errors in their budgeting process, Azure AI identified outliers so they could reclassify the spending and rectify the errors.

• Modeling and assumptions.

  Forrester modeled the impact for the composite organization assuming:

• The composite applies ML to help a team of eight data scientists and engineers.

• Your Organizationmight apply ML to help a team of x data scientists and engineers.

• Azure AI reduces the effort required by 15% in Year 1, and 25% in Years 2 and 3.
• Data scientists and engineers have fully burdened annual salaries of about $156,940, or $75 per hour, on average.
• The organization processed 2 million pages per year for customer work before the investment in Azure AI and had a document error rate of 4%.

• Your Organization might process x tasks/transactions per year with an error rate of x.

• With Azure AI, the organization reduces their document error rate by 60% with Azure AI.
• On average, customer success employees spends 20 minutes dealing with each error.
• Customer success managers have fully burdened annual salaries of about $93,000, or $45 per hour, on average.
• Each FTE recaptures 50% of their saved time into productivity value.

• Risks.

  The expected financial impact is subject to risks and variation based on factors including the organization’s:

• Azure AI scope of deployment and use cases.
• Level of previous investment in internal AI or ML workstreams.
• Previous number of documents processed and document error rate, as well as the consequences of errors and time spent fixing them.
• Level of internal investment in model development and fine tuning.
• Number of project team members and their compensation amounts and structures.
• Unique business requirements and complexities that may reduce potential time savings.

• Results.

  To account for these risks, Forrester adjusted this benefit downward by 15%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $1 million.

• For Your Organization , the three-year, risk-adjusted total PV might be x. (To account for risks and based on your planned Azure AI usage, this benefit is adjusted downward by x.

Retired Legacy Technology

• Evidence and data.

  The interviewee’s organizations decommissioned legacy on-premises and third-party point solutions in favor of Microsoft Azure’s pay-for-what-you consume cloud licensing model. In making this investment, organizations realized savings from legacy licensing, maintenance, administration, and support costs.

• The partner at the professional services organization described realizing savings by investing in Azure AI in data infrastructure, software, and the labor associated with integration and ongoing maintenance. They said: “At the earlier stages we had our own servers and open source [tools] with Python code and different frameworks to connect the front and back end, which was super manual. The maintenance of all those stacks, and budgeting [them] was super complex. … Working with a hyperscaler [solves for that].”
• The program manager in retail mentioned a major transformation of their organization’s entire data landscape into Azure cloud pipelines and a data lake. They described, “[Now] we pay for a service, and for consumption, but it’s much easier to manage. Before, you needed an entire team of accountants just to manage the deprecation on the technologies that we had.”
• The senior principal of digital at the aviation firm said: “We don’t have to maintain the OCR infrastructure anymore. It’s a Microsoft problem. It allowed [engineers] to focus more on [other things]. It took them an average of two hours per week per engineer to troubleshoot issues. Now we are down to zero because we offload [the work] to Azure.”

• Modeling and assumptions.

  Forrester modeled the impact for the composite organization assuming:

• The composite previously spends $42,000 annually on on-premises data infrastructure and licensing.

• Your Organizationmight spend x annually on legacy on-premises data infrastructure and licensing.

• The composite previously spends $10,500 in labor to support the on-premises infrastructure.

• Your Organization might spend x annually on labor to support legacy on-premises infrastructure.

• The composite, upon moving to cloud-based Azure AI, retiring 20% of the previous technology in Year 1, 40% in Year 2, and 85% in Year 3.

• Risks.

  The expected financial impact is subject to risks and variation based on the organization’s:

• Previous data infrastructure and associated software and maintenance costs and its ability to retire solutions after the investment in Azure AI.
• A percentage of legacy solutions were decommissioned, and a percentage of on-premises solutions moved to the cloud.

• Results.

  To account for these risks, Forrester adjusted this benefit downward by 5%, yielding a three-year, risk-adjusted total PV (discounted at 10%) of $57,000.

• For Your Organization , the three-year, risk-adjusted total PV might be x. (To account for risks and based on your planned Azure AI usage, this benefit is adjusted downward by x.)

Unquantified Benefits

Interviewees mentioned the following additional benefits that their organizations experienced but were not able to quantify:

• Optimized budgeting and reallocated spending leading to greater returns

  The principal AI architect at the energy organization described that decision-makers used cost models and revenue projections from Azure AI to get the best capital efficiency possible: “[Decision-makers] can really deploy [their budgets] for commercial value, because the variance between the accuracy of forecasts and what actually happens is now dramatically reduced.”

• Faster and higher quality decision making.

  The partner in the professional services organization described, “We went from getting no answers to getting answers in a very good timeframe.” The retail program manager also mentioned that predictions from Azure AI influenced decisions at the executive board level.

• Greater innovation and cross-team alignment.

  The same program manager told Forrester: “[Azure AI has] given us a data-driven mindset, which has gone across multiple other use cases now.” They described that having some successful use cases from the start to demonstrate value helped AI technology gain credibility with the business side of the organization as a true unlocker of business potential. They said, “It’s a way to get [the IT and business] teams to work as one team.”

• Improved employee safety.

  The principal AI architect at the energy organization mentioned that enabling remote equipment operations and monitoring removed significant human safety risks from employees or contractors previously flying out to remote places to inspect heavy machinery in person. The organization also used Azure Custom Vision to ensure safety standards were abided. They said, “With things like Azure Custom Vision, you’re identifying and classifying objects and in real time, making risk assessments that then get flagged up to [take] action, saying, [for example]: ‘because we’re doing this production process, we shouldn’t have people in that area because they don’t have the right protective equipment.”’

• Enhanced model compliance and governance, and insurance of responsible AI/ML practices.

  Microsoft espouses a high commitment to responsible AI development and advancement. Azure AI can help organizations quantify model fairness, robustness, and security, and ensure that their models are governed properly and adhere to ethical principles.

• Reduced greenhouse gas emissions.

  The principal AI architect at the energy organization highlighted that by reducing fuel use of shipping vessels, Azure AI was helping the organization get closer to its goal of reaching net zero emissions.

• Improved EX.

  By releasing employees from routine, manual tasks, employees can spend a higher percentage of their time on more engaging work. Azure AI as a highly-valued technology may also be beneficial for recruiting. The principal AI architect in energy noted, “[Azure AI is] one of our major selling points when [our engineers] go out and recruit graduates in the market.”

• Reduced risk.

  AI helps organizations more effectively identify and manage risk and reduce the risk of compliance violations, penalties, or fines. Similarly, predictive analytics use cases allow anticipation for future events such as equipment failure or customer behavior, enabling decision-makers to proactively take action to mitigate risk. For example, the principal architect at the energy organization said their organization used Azure custom vision to identify people smoking at flammable, high-risk areas like gas stations or near gas pipelines.

Flexibility

Flexibility represents additional capabilities that could be turned into future business benefit, providing an organization with the ability to engage in future initiatives. The value of flexibility is unique to each customer. Scenarios in which a customer might implement Azure AI and later realize additional uses and business opportunities include:

• Further operationalizing AI across an organization.

  As AI becomes more affordable, it is expanding beyond standard use cases to be used by roles beyond data scientists to provide value across most parts of a business. The principal architect of AI in energy said, “Just every use case you might think of, retail or finance or insurance, effectively, [AI is] part of some business and is being applied.” Organizations can continue to increase the value of their AI development by building off existing applications.

• Improved innovation, strategy, and productivity by de-risking AI and ML.

  The same technological progress that has made AI and ML more affordable and accessible across organizations spurs innovation by allowing more roles to think creatively about its usage. The program manager at the retail organization described, “[Azure AI has] allowed us to take away the guardrails of financing being a constraint and allows people to come [forward] with an idea.” The principal AI architect at the energy organization echoed: “There’s an opportunity to test the value very quickly and see if it’s there before making the big investments that you might spend four or six months doing to build an apps or something that could be globally deployed.”

• Capitalization on cutting-edge technological innovation.

  Organizations with institutional knowledge and an established AI or ML team working with Azure AI can realize value from new products or advances in technological capabilities much sooner than others. The partner at the professional services organization noted that their partnership with Microsoft allowed them to access new products, “We are continuously bonding with Microsoft and getting support on the next products in terms of AI and also ML.”

Flexibility would also be quantified when evaluated as part of a specific project (described in more detail in Appendix A).