WHY AISWITCH
Why AISWITCH
AI-automation Market Predicts:
- 73 million jobs are to be automated by 2030 (McKinsey)
- AI would create net net 500,000 jobs by 2020 and 2 million by 2025 (Gartner)
- The market for AI & automation technologies is predicted to grow at 25-30%+ CAGR, next decade.
- By 2025,15-trillion+ USD of global economy may be generated/ powered by AI-automation.
- More than 50% jobs will be made redundant by AI-automation, within next 3-5 years(Quartz- 80+ countries)
- A 2019 survey in US showed only 29% companies use AI on a regular basis.
- 75%+ adults in US believe that AI will eliminate more jobs than it creates. (Gallup survey)
- By 2025, AI will help manufacturing sector to reduce waste by up to 50% (WEF)
- Annual investment in AI startups at $26.6 billion in 2019, up from $4.2 billion in 2014
- 2020 onwards, it will take more than 17 years for AI to generate more jobs than it replaces, in UK (PWC)
AI sunk costs: AI is failing to scale, in 90%+ enterprises
Gartner said in 2020 AI hype cycle- AI governance is fast emerging as the biggest headache.
All net new digital businesses are AI-native, e.g. digital banks, blockchains, cryptocurrency companies, new-age insurance companies like Lemonade, digital advertising, legal processing, and post-COVID, digital healthcare services spectrum- remote monitoring to drastically faster clinical trials in pharma.
But, AI ground realities faced by traditional enterprises are quite grim. Even the early adopters are facing challenges in adoption@scale. Without adoption, RoI is low, and payback periods for these expensive tech-stacks are beyond 3-5 years. Consequently, business sponsorship and budgets for these technologies are drying up under duress, especially when decisions are taken in a short to medium-term outlook.
WHY DOES AI FAIL?
- Strategy gaps- No consistent strategy & policies
- Workforce issues - Skills not available
- Information- Data quantity vs. quality, security, bias, relevance, 4 V's
- Technology- No reliable guidance on AI tech-stacks & costs
- Culture change- biggest hurdle to adoption & scale
- Human-AI augmentation metrics- No clear outcomes
Industry research on AI is shady and murky, with unverifiable and doubtful survey-sourced data, and pseudo-experts camouflaging as AI tech/ practice SMEs, e.g. a global newspapers' misleading marketing claim of a fully GPT3-generated article! Given the vastness, depth, complexity, change dynamics and costs of AI-automation tech-stacks, it is absolutely imperative to have evidence-based, actionable frameworks, guidance & benchmark data on cost, accuracy, performance. AISWITCH fills these gaps, with 2 key solutions:
- AI-automation focussed, evidence and practical experience-based research, and verifiable, 100% traceable data & benchmarks
- Patented, ITIL-like frameworks, to manage 6 key enterprise AI dimensions, in AI-SWITCH (IP): AI Strategy: Cost, policy, risk; Workforce: CoE, reskilling, KRA-KPIs; Information: data quality, bias, 4V's; Technology: AI-optimized, green infra, algo's; Culture: Change Islands, speculative design, CUDA; Human-AI augmentation - outcomes & value metrics
Top 3 practical challenges in AI & automation
- To realize targeted RoI, companies will have to scale up AI and RPA applications to end-to-end processes: Current state of adoption in most enterprises looks like this: 25-50 bots built on a few RPA and AI tech-stacks, automating specific tasks such as image classification or document-processing, as point solutions. At this rate, RoI from these stacks will be a far cry from what the vendors and tech partners promise.
- There's no standardized way to manage these assets@scale, with consistent processes: Scaling up @prod means - 1000's of bots running in integrated process chains. But the AI & RPA tech-stacks run on their own orchestrators and runbooks/ playbooks only.
- Already there are heavy sunk costs i.e. high capex spent in setting up & licensing expensive tools/ tech, be it SaaS or on-prem. Then there is even higher opex: Due to extreme lack of consistent processes across the tools-stacks, AI-RPA users suffer from spiralling 'hidden' opex.
Case in point
To make simple changes in decision variable-values/ thresholds, in a bot's decision-tree/ logic layer/ automated workflow, some vendors and service partners demand months of efforts and charge a bomb, inflating the timelines and hiking costs, lock-in & tech debts.
- Should change execution in RPA be expensive? Didn't these tools promise to drastically reduce the need for deep-tech/ code skills, and give controls to business users, with low-code/ no-code options?
- Even for changes in AI solutions e.g. in datasets/ decision variables/ hyperparameters/ algorithms/ models, how are the efforts estimated and priced?
Multiply these mini-projects' cost-spikes with 1000's of RPA bots & AI-powered agents running in integrated/ cascading workflows, to enable end-to-end process automation. You end up hitting a steep cost-wall!